Possible fauly memory

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View
Friend has a home built computer; Asus P5N-E SLI Motherboard V Dual Channel
kit 2x1gb memory V Intel Core 2 Duo E6550 2.33Ghz 1333Mhz Socket T (LGA775)
CPU. Running on Vista Home.

For a long time she has experienced intermittent computer freezes,
intermittent BSOD  (minidump shows many different causes) and corrupt
images made with Acronis TrueImage.

I recently tested the memory using Memtest86+ v4.20 and the test indicated
problems. Purchased new dual kit 2x1gb memory, removed the old memory and
installed the new. Computer boots up okay and the machine runs normally,
but when I ran a test with Memtest86+ on the new memory it again came up
with problems.

Im no expert but to my thinking this would indicate a faulty motherboard.
I mentioned this to her and she mentioned that the problem was evident with
the same memory problem when the computer was originally built  and at the
time the builder changed both the memory and the motherboard.

I intend removing one stick of memory and then retesting V this would
indicate if the memory is faulty, and if both sticks come up faulty this
would indicate that the motherboard was faulty. Am I thinking along the
correct lines or is there something else I should test for?  
Any suggestions would be most appreciated.

Re: Possible fauly memory

On 6/10/2011 6:36 PM, John wrote:
Quoted text here. Click to load it
As long as your Memtest86+ came from a bootable CD or Flash memory and
it was created on a different computer then I tend to agree with you but
there still might be another cause.

My home built system using a XFX MG-63MI-7109 motherboard also reported
unstable memory when using the factory default BIOS settings and memory
test.  I found that if I kept the default BIOS memory timing values but
raised the memory voltage a bit that everything became stable.  On mine
it offered color coded voltage settings which I used to keep the voltage
below what the motherboard considered excessive but still higher than
the default values it was set at.  I seem to remember going from 1.8 to
1.94 volts but it's been more than a year since I looked at the BIOS

Not sure if her problem motherboard BIOS offers options on memory timing
and voltages but it might be worth looking into.

Re: Possible fauly memory

On Fri, 10 Jun 2011 20:40:31 -0500, GlowingBlueMist wrote:

Quoted text here. Click to load it

I created the CD on my computer.
Quoted text here. Click to load it

When I installed the new memory I had numerous problems so changed the
memory voltage to 2.178v in the Bios and everything worked great, apart
from Memtest86+ telling me different:-) The Bios setting were all set to
auto before changing to 2.178v, and this was the only setting I changed,
maybe doing further changes might be the answer. Thanks for the feedback:-)

Re: Possible fauly memory

John wrote:
Quoted text here. Click to load it

I think you're on the right track. I would test one stick of memory,
and test it in both channels (two separate test cases). The idea is,
to see if one of the memory channels is defective.

In some cases, it's one of four slots which is bad. So you can
spend a fair amount of time running test cases, to isolate the
type of fault.

Occasionally, there'll be a motherboard, where the BIOS doesn't actually
set the memory parameters which are displayed in the BIOS setup screen.
It helps in that case (at least if you're a Windows user), to grab a
copy of CPUZ, and double check while in Windows, what has been set up.
I have a motherboard here, that doesn't have agreement between the
timings shown in Windows by CPUZ, versus what the BIOS shows. This is
a BIOS bug.

Nvidia chipsets have had issues of their own. Typically, showing up
as an inability to run four DIMMs, at the stated chipset speed. To check
for that, you go to the appropriate Asus forum and see what others think
of the product.

(Arrgh! This thing is huge... Too many posts. And the search options here, suck.
But if you need to do data mining, this is the place to go.)


On the first page of hits there, I can see someone with a bad DIMM slot.
Presumably a logic signal failure, as there is no claim they were moving
memory around at the time.

Another thing to be aware of, is a couple Nvidia chipsets are
sensitive to ESD (static discharge) on the PCI Express slot pins.
Knowing that, if you're going to work on that system, for peace of
mind, use a wrist strap. One end goes on your wrist, the other to
the chassis, and any removed items go in antistatic bags etc. The
wrist strap has one thing in it you cannot see. There is normally
a 1 megohm resistor in series with the strap. The strap is not
just a "piece of wire" and a piece of wire is not an exact substitute.
The 1 megohm to 22 megohm series resistor is there to drain static
slowly, over a fraction of a second. That avoids a large current peak.
If you want to use a piece of wire, add a resistor in series.


Normally, silicon chips are resistant to ESD to a 1000 or 2000 volt
level. That makes the chipset resistant to normal handling. Some
USB ports are rated at 5000 to 6000 volts. But if a chip designer
is having trouble making an I/O signal operate at Gigahertz
speed, sometimes the only option, is to fiddle with the ESD
network (change the structure). And doing so, means somebody
else could be faced with an overly sensitive product to work with.
The only reason I suspect something like that, is the number
of reports of blown out video card slots. An ESD strap should
give some measure of protection. Other designs are resistant
enough, that causal attention to ESD mitigation is enough.

I've worked with one silicon chip, which due to a design oversight,
had no (zero) ESD protection. The chip was so sensitive, 50%
mortality results from the chips sliding down the *inside* of
an antistatic tube :-) That's what happens when there is no
protection. The NVidia chip wouldn't be that bad, because
the motherboards managed to leave the factory.

One other bug you might want to know about. I've seen a
thread before, where disk corruption was occurring, and
the corruption was a limited number of bytes (like four wrong
bytes in say a 16KB block, a 32 bit word corrupted). There seemed
to be some relationship between the NVidia built-in NIC in the
Southbridge, and the disk corruption issue. I never did see a successful
resolution, and no claim that a BIOS change fixed it. Since
you have evidence of a memory problem, via memtest, there is
no need at the moment to entertain other sources of trouble.
But there was at least one other way to get disk corruption,
and it seemed to involve an issue with the internal busses in
the chipset. In such cases, if the motherboard has two LAN interfaces
(one Nvidia, one an external LAN chip of another brand), you select
the non-Nvidia interface and re-test. Or, add a LAN card and hope
the issue doesn't also exist when the PCI bus is used.


Re: Possible fauly memory

John wrote:
Quoted text here. Click to load it

IOW, the memory didn't work, and unless this is a BIOS problem, it
probably means at least one memory chip is faulty.

Quoted text here. Click to load it

In general, higher priced mobos like yours are much less likely to be
faulty than the memory, and such mobos are unlikely to degrade in a
few years due to rotting capacitors because they're made with good
capacitors.  OTOH even most major brand, lifetime warranted, retail
memory modules are made with no-name chips, much of it untested or
rejected by the actual manufacturer.  One sign of low quality is a
voltage rating above 1.8V (all prime quality DDR2 chips pass all
factory testing at 1.8V); another is the lack of visible markings
identifying the brand or part number.  So according to what you can
see on the modules, who made yoru memory chips? (Nanya, Hynix, Micron,
Samsung, etc.)

Re: Possible fauly memory

On Sat, 11 Jun 2011 11:36:17 +1200, John wrote:

Quoted text here. Click to load it

First off I would like to thank everyone that replied to my posting, all
replies have been copied and saved.

Since I first posted my problem with memory on my friends computer I've
spent sometime reading up on memory problems (most of which went over my
head)but there are a few points that seemed interesting;

* With Asus motherboards you need to disable legacy usb in the bios for
Memtest86+ to work correctly - has anyone heard about this and if so, is it

* Reduce memory bus speed. How would I do this?

*With memory timings is it just a simple task of changing the bios as this
setting shown here? CAS-tRCD-tRP-tRAS 2-2-2-5 - i.e. CAS 2 tRCD 2 tRP 2
tRAS 5?

It will be a few days before I have access to the computer so the more
information I get now the better:-)


Re: Possible fauly memory

John wrote:
Quoted text here. Click to load it

It was true at one time. That may not have been an issue with all motherboards,
but a select few. Both the motherboards, and memtest, have changed since

Memtest, relies on a certain standard, for reserved memory areas.
Memtest won't test the very lowest memory, because the BIOS may still be
using it. Memtest relies on the information the BIOS gives it, to not
cause a conflict.

Quoted text here. Click to load it

1) Download the user manual (PDF format) from support.asus.com

2) Change "Ai Tuning" to [Manual]. That should expose more settings for
    manual adjustment. I usually spend 30 minutes going through a BIOS,
    setting things to manual, and seeing what extra settings get added
    to the list.

3) See "FSB and Mem Config" in the manual, to adjust the memory clock.
    "Linked" mode, implies standard ratios between FSB and memory clock.
    "FSB - Memory Ratio" applies a particular divider.

    For an example, say you're using a FSB1066 processor at stock.
    The FSB is a QDR interface or quad data rate.
    Divide by 4 to get the actual CPU input clock.
    1066/4 = 266MHz CPU input clock. Now, use a memory divider of 3:2.
    That gives 266*3/2=400. Multiply that memory clock by 2 (double data
    rate memory type) to get a DDR memory rate of DDR2-800 (PC2-6400).
    If you went from 3:2 down to 5:4, the math gives DDR2-667 (~ PC2-5300).

4) If you use that control, you can leave the timings at Auto.
    If you crank the CPU clock manually, that throws off the BIOS
    calculation of timing values, and you should do the math yourself
    and scale them, according to the final memory clock you've set up
    for yourself. If the clock is at stock (like 266Mhz for FSB1066),
    then the BIOS should be able to interpolate the correct values, out
    of the values listed in the memory DIMM SPD EEPROM chip.

5) There is also a Command Rate setting. 2N is more relaxed than 1N,
    and limits address transmission to every second cycle. On
    older systems, it was only needed for four DIMMs. On modern systems
    now, with sky-high memory clocks, 2N is needed even with one DIMM
    per channel.

    In the manual, look for "Command Per Clock" and test with [2 Clock]
    and see if stability improves.

6) In terms of timing adjustments, if the CAS was 5, that is five
    clock ticks. Each clock tick on the memory, has a period in nanoseconds.
    So when you talk of 5 ticks, you could be speaking of 12.5 ns worth
    of time. When the memory input clock is changed, the 12.5ns parameter
    you're trying to meet, doesn't change. You still want to request
    data to arrive 12.5ns later. Say the new clock period was 2 nanoseconds
    (i.e. higher clock). 12.5/2 = 6 plus a little bit, which you round
    up to 7. So you'd go into the BIOS, and change CAS from 5 to 7, if you'd
    raised the clock. You can dispense with a traceable calculation, and
    just scale the numbers by the ratio of the clocks (if you can figure
    that out). For example, say by playing with just the CPU input clock
    (i.e. overclocking the CPU), I'd actually forced the memory from
    DDR2-667 to DDR2-800 and CAS was 5 at 667 (said so on the packaging
    the memory came in). I could guess at the new CAS value as 5*800/667
    or 6 exactly (since there is no fractional part, there is no need to
    round up).

    But in your case, you're not changing the CPU input clock from the
    stock value. And if you do that, the BIOS should calculate the timing
    properly on its own. You can use CPUZ while in Windows, to verify the

    I should note, that this isn't always true. The BIOS being good at it :-)
    The computer I'm typing this on, doesn't know how to scale properly,
    and ran my RAM with a CAS which was too tight. And the RAM still
    passed :-) But when I checked CPUZ, that really pissed me off, because
    I prefer to be in control. I don't want my hardware trashed, because
    some BIOS designer couldn't do their job. At the time, I wasn't trying
    to overclock, but discovered soon after that the BIOS had cranked stuff
    on me, and not corrected the CAS. According to the specs, the computer
    should have crashed immediately.

Quoted text here. Click to load it

The DDR2 memory, will be using higher values than that. I've got DDR2
in this machine, and I own a pair of CAS4 and a pair of CAS5 memories.
Like perhaps 5-5-5-18. So expect to see higher than 2-2-2-5, which is
an older, enthusiast DDR400 spec. (On some of the DDR3 stuff, with
sky high clocks, they're up around CAS 9.)

Quoted text here. Click to load it

Plenty of "fun" ahead.


Re: Possible fauly memory

On Sun, 12 Jun 2011 04:41:13 -0400, Paul wrote:


Thanks Paul, you've given me plenty to think about.

Quoted text here. Click to load it

So is it a simple case of entering the number i.e. CAS 5 tRCD 5 tRD 5 tRAS
18 in that order (assuming that 5-5-5-18 is the memory settings for her

Re: Possible fauly memory

John wrote:
Quoted text here. Click to load it

Yes, you can do that. They're integer values on modern memories.
The DDR400 memory, supported CAS2, CAS2.5, CAS3, but more recent
memories use integers like CAS4, CAS5, ... CAS9. On DDR, CAS could
be launched on either edge, while I'm guessing the higher speed
memories chose to only launch on one edge.

When in the BIOS, if you cursor down to the field you want to
change, the instructions for changing it should be in the upper
right hand corner of the BIOS screen. It'll either say to use the
+/- keys, or key in the value directly. Those descriptions can
be pretty cryptic, so it isn't always easy to figure out what
they're saying. And the manual may not expand or offer any more
useful info.

If you're totally stuck, for what the BIOS settings do, you can
try this site.



Re: Possible fauly memory

On Sun, 12 Jun 2011 16:11:01 -0400, Paul wrote:

Quoted text here. Click to load it

I've bookmarked that site and intend spending a few hours browsing:-)

Thank you again, Paul.

Site Timeline