[plug] Memory testing

Arie Hol arie99 at ozemail.com.au
Wed Mar 5 22:57:14 WST 2003


At 10:12 PM 05-03-2003 +0800, you wrote:
>Craig
>
>Have tried many varied combinations of three slots for RAM.
>Last config I tried was 128 in Bank 0 and 256 in Bank 2, these numbers
>are those seen in BIOS.
>
>Error:
>
>CPU 0: Machine Check Exception: 0000000000000004
>Bank 2: f60020000000017a at 000000000be5f000
>Kernel panic: CPU context corrupt
>Mar 5 21:15:47 Coven kernel: CPU 0: Machine Check Exception: 0000000000000004
>Mar 5 21:15:47 Coven kernel: Bank 2: f60020000000017a at 000000000be5f000
>Mar 5 21:15:47 Coven kernel: Kernel panic: CPU context corrupt

1 - Are you using DIMM's or SIMM'S ?

2 - Are they single or double sided ?

3 - Or are you using one of each ? - If so check your m/board manual on
using mixed RAM configuration.

4 - Are they parity/ECC - nonparity/NON-ECC ?

5 - Are they both same brand and speed ?

M/boards with only three RAM slots often have a restriction on the use of
mixed DIMM's/SIMM's. There may be a conflict as to which bank will be
assigned to each DIMM/SIMM. This conflict may be brought about due to the
high load you are placing on your system when compiling.

Generally - three slot m/boards can use :

a) two doublesided DIMM's/SIMM's assign RAM (slot 1 = bank0 & bank1 : slot 2
= bank2 & bank3)

OR 

b) three single sided DIMM's/SIMM's assign RAM (slot 1 = bank0 : slot 2 =
bank1 : slot = bank2)

OR

c) one doublesided DIMM/SIMM in slot 1 assigns RAM (slot 1 = bank0 & bank1 )
with a single sided DIMM/SIMM in slot 2 assigns RAM ( slot 2 = bank2) -
Might work if your lucky !!!

OR 

d) one single sided DIMM/SIMM in slot 1 assigns RAM ( slot 1 = bank0) with
one doublesided DIMM/SIMM in slot 2 assigns RAM (slot 2 = bank2 & bank3 )
this "shouldn't work" because there is NO bank1 !!!

I have a technical reference on RAM somewhere but haven't seen it for ages.

Regards Arie

>--------------------------------------------<

For the concert of life, nobody has a program.

>--------------------------------------------<



More information about the plug mailing list