Question on DDR Ram problem

[Guest]

Archived from groups: comp.sys.ibm.pc.hardware.chips (More info?)

Hi, I have a funny problem on my eMachines (FIC AM39L motherboard, Via
KM400A chipset) with memory.

I recently got 512MB Kingston memory to add to my 512MB Samsung memory
that came with the system.

I put the Kingston in the first slot, and the Samsung in the second, and
my system was flaky.

I switched them, and my system was stable, but I wanted to make sure and
got MEMTEST-86.

With 1 gig installed, I am getting a few errors on test 5 (never the
same amount, and never at the same place).

With either chip in the first slot, I get no errors at all!!!!!

Is there a way I can diagnose what is going on? Could the extra memory
be dragging down the amount of power causing the problem?

Or what else could be the problem? The system seems stable enough now,
but I am not very comfortable with MEMTEST-86 failing (only on test 5).

Thanks for any advice,
Larry

 

[Ed]

Archived from groups: comp.sys.ibm.pc.hardware.chips (More info?)

On Wed, 01 Dec 2004 13:51:58 -0500, Lawrence Guros
<Lawrence.Guros@oracle.com> wrote:

>Hi, I have a funny problem on my eMachines (FIC AM39L motherboard, Via
>KM400A chipset) with memory.
>
>I recently got 512MB Kingston memory to add to my 512MB Samsung memory
>that came with the system.
>
>I put the Kingston in the first slot, and the Samsung in the second, and
>my system was flaky.
>
>I switched them, and my system was stable, but I wanted to make sure and
>got MEMTEST-86.
>
>With 1 gig installed, I am getting a few errors on test 5 (never the
>same amount, and never at the same place).
>
>With either chip in the first slot, I get no errors at all!!!!!
>
>Is there a way I can diagnose what is going on? Could the extra memory
>be dragging down the amount of power causing the problem?
>
>Or what else could be the problem? The system seems stable enough now,
>but I am not very comfortable with MEMTEST-86 failing (only on test 5).
>
>Thanks for any advice,
>Larry

If the 2 sticks of ram don't have the same timings you'll probably need
to set them manually in the BIOS to the slower stick of ram.
Common settings...
CAS Latency
RAS to CAS Delay
RAS PreCharge Delay
Active PreCharge Delay
Command Rate

Memory Basics
http://www.corsairmicro.com/memory_basics/153707/index.html

FIC AM39L motherboard manual
ftp://ftp.fica.com/manuals/motherboards/Socket%20A/AM39L/

Ed

 

[Guest]

Archived from groups: comp.sys.ibm.pc.hardware.chips (More info?)

On Wed, 01 Dec 2004 13:51:58 -0500, Lawrence Guros
<Lawrence.Guros@oracle.com> wrote:

>Hi, I have a funny problem on my eMachines (FIC AM39L motherboard, Via
>KM400A chipset) with memory.
>
>I recently got 512MB Kingston memory to add to my 512MB Samsung memory
>that came with the system.
>
>I put the Kingston in the first slot, and the Samsung in the second, and
>my system was flaky.

All DRAM from different mfrs has slightly different tolerances on timings -
especially with large amounts, like 1GB, it's a bad idea to mix. Do you
know the timings on the Samsung?

>I switched them, and my system was stable, but I wanted to make sure and
>got MEMTEST-86.

In your BIOS Setup, are the timings all set to use SPD? A good BIOS should
examine SPD of all modules and use the slowest combined timings, or if the
chipset supports separate individual timings apply them... but many don't
get it quite right. Some will even report the values on the POST screen.

>With 1 gig installed, I am getting a few errors on test 5 (never the
>same amount, and never at the same place).

I've seen a few failures on test 5 when timings were too tight - start by
pushing up tRCD and then CL. Get memtest86+ from www.memtest.org and it'll
show the actual timings being used. The conventional order is CL - tRCD -
tRP - tRAS - xT, typically 3-3-3-8-2T. From what I've seen, the command
rate, 1T or 2T, has the most effect on bandwidth performance on current
DIMMs.

>With either chip in the first slot, I get no errors at all!!!!!
>
>Is there a way I can diagnose what is going on? Could the extra memory
>be dragging down the amount of power causing the problem?

Yep that's likely the problem - the memory channel is more "loaded". It
could be that your different DIMMs require different voltages too -
Kingston seems to specify higher, up to 2.7V - for their faster DIMMs. If
you can adjust voltage in BIOS try bumping that, to a max of say 2.75V, as
well as relaxing timings.

>Or what else could be the problem? The system seems stable enough now,
>but I am not very comfortable with MEMTEST-86 failing (only on test 5).

That's not unusual. Get Prime95, from http://www.mersenne.org/freesoft.htm
which runs under Windows and run the "torture test"... for a good hour or
more. It often finds errors when memtest86 or memtest86+ has not - in my
testing this happened with command rate at 1T and pushing it to 2T "fixed"
things.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[Guest]

Archived from groups: comp.sys.ibm.pc.hardware.chips (More info?)

Lawrence Guros <Lawrence.Guros@oracle.com> wrote:
> Hi, I have a funny problem on my eMachines (FIC AM39L
> motherboard, Via KM400A chipset) with memory.
[snip]
> I switched them, and my system was stable,

Very typical for different memory ratings. AFAICS, the
chipset only reads one DIMM EEPROM for it's settings.

> Is there a way I can diagnose what is going on? Could the
> extra memory be dragging down the amount of power causing
> the problem?

Entirely possible that it is bus loading. First try
manually changing all memory settings to their slowest.
If memtest is clean, reset half faster. If not, replacing
the PSU _might_ help, but most likely it is a poor mobo or
mem modules.

-- Robert