kandinsky63 said:
Is there a convention that applies to size and brands of RAM modules
when installing them?
For e.g. if there are 2 DIMM sockets is it best practice to
1. add 2 ram modules of the same capacity and same manufacture OR
2. can the selection be mix and match i.e.
a.) 1GB in DIMM 1 and 512MB in DIMM 2 from the same
manufacturer
b.) 1GB in DIMM 1 and 512MB in DIMM 2 from different
manufacturer?
My current ram config is 2 x 512MB of Brand B
I decided to install 1GB of Brand A in DIMM 1 and left 512MB of Brand B in
DIMM 2. I then ran a quick memtest84+ on this config which detected no
errors. However, when I booted up and left the system running for about 1
hour, I got a blue screen with the error message 'problem caused by the file
'Ntfs.sys' PAGE_FAULT_IN_NONPAGED_AREA.
My only conclusion is that there may be an incompatibilty between either two
different sized RAM modules or between the 2 different brands.
When running memtest84+ on the 1GB ram Brand A on its own , no errors where
found.
I would appreciate some input on this issue.
I wonder if you're using "high density" 1GB DDR RAM modules ?
Did you buy the 1GB modules from Ebay ? There are some sellers
there, who sell the high density RAM, and the advert will also
mention the fact they are high density. I would recommend
against the high density type, and get a more conventional
RAM. (High density uses x4 or four bit wide chips, while
regular memory uses x8 or eight bit wide chips arranged
in two ranks. While there are chipsets that support both,
theoretically speaking, the x8 modules are a better choice.
It isn't clear how many high density modules will run on
a motherboard, without problems.) High density RAM is
mainly found on Ebay, because real retailers don't want any
part of it.
To accelerate memory testing, I recommended a few different
test cases.
1) Use Memtest86+ for a quick initial test. Memtest86+ is
for weeding out severely compromised setups, ones that
aren't even close to stable. You use Memtest86+ as a check,
to avoid corrupting your Windows install by booting up with
really bad RAM.
2) If you can complete a couple passes of Memtest86+, your
next test is to use Prime95 from mersenne.org . I've
had memory that was so bad, it failed in about ten
seconds of testing with Prime95. I generally use four
hours of error free testing, as proof it is stable
enough to use for real work.
3) Mixing in some 3D video work, along with Prime95, can
be used as an extension of (2). For example, I have some
RAM here right now, which passes (2), but fails if I play
a 3D game while Prime95 is running.
The purpose of (2), is to quickly conclude whether the
RAM and setup are correct or not. An hour of Prime95
torture testing, is equivalent to about 100 hours of just letting
the system sit there displaying the desktop. So the Prime95
Torture Test speeds up testing immensely.
As for your memory settings, your Asrock P4VM800 doesn't have
a lot of options. With two sticks of RAM, changing to Command
Rate 2T can help you. Reducing the memory speed, below the
current value may also help. The purpose of dropping the
speed, is to see if you can make the RAM stable under
any circumstances at all.
I'm facing a similar issue right now, with my Asrock product.
My computer seems stable enough with 1GB DDR2 modules, and
I've been experimenting with some new 2GB modules. Something
obviously needs to be tuned (because I get the occasional error when
using two sticks, but no errors with only one stick). I
suspect it is just another symptom of a half finished
BIOS design. There are some RAM parameters, which are not
shown in the BIOS screen. And in some cases, as a user, you
don't know which direction to move the controls, to fix it.
For example, some boards have "clock skew" controls, and
it can takes hours of experimenting to figure out what
they do. Back in the lab, we'd use a high speed scope, to see
what is going on, but for home usage, that isn't an option.
Increasing the CAS value, is less stressful for the RAM.
Increasing Command Rate to 2T is less scressful.
Decreasing memory clock speed 200-->166-->133MHz, is less stressful.
Enabling Flexibility mode is not explained fully, but seems
to reduce the speed.
There is also a DRAM Voltage option, which is set at
Auto by default. Since Asrock chooses not to
calibrate their settings, instead offering "high",
"low", and the like, we cannot really guess as to
what a correct setting might be. Many other manufacturers
list the actual voltage that should result, like 2.6V
or 2.7V and so on. 2.7V or 2.75V should be safe with
most brands of DDR RAM.
Paul
