LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Linux - Hardware (https://www.linuxquestions.org/questions/linux-hardware-18/)
-   -   added ram - system unstable, segfaults (https://www.linuxquestions.org/questions/linux-hardware-18/added-ram-system-unstable-segfaults-379331/)

slackhack 11-02-2005 01:29 PM

added ram - system unstable, segfaults
 
i just added 512 PC3200 to my 2x256MB PC3200, and it's giving me major problems.

motherboard is nforce2, and i'm not overclocking anything right now. the memory is coincidentally the same brand as the 2x256, all buffalo tech, pc3200 cas 2.5.

i had the 2x256 set at the motherboard's "auto" and "turbo" settings (2.5-2-2-6), it said dual channel DDR400 when it booted so i guess it was running dual channel okay, and it gave me no problems at all. i could overclock the fsb to 200 and it was fine.

with the 512 stick added, at turbo settings it crashes all my applications with segfaults. i read something on LQ about specifically setting the ram speed instead of leaving it on auto if things aren't working, but if i change the ddr % setting even to to 120% it won't post (which at 166mhz is less than ddr400, is that right? i can't even get the rated speed?). i tried raising the cas to 3 (3-3-3-7) but it also wouldn't post. the "optimal" setting (2.5-3-3-7) boots (which i completely don't get, since it won't boot with the exact same settings except for cas 3) but it's a little unstable (screensaver coming on out of the blue, etc.) when i kill X there are all kinds of screensaver and other errors that say something about "undefined symbols" i think.

all the memory doesn't show up in free, either:
Code:

[6] sero:~ $ free -o
            total      used      free    shared    buffers    cached
Mem:        906316    154984    751332          0      9480      85984
Swap:      987988          0    987988

[7] sero:~ $ free -om
            total      used      free    shared    buffers    cached
Mem:          885        151        733          0          9        83
Swap:          964          0        964

where's the rest of the MBs? :confused: it's detected at post, but not in linux -- why would that be?

i ran memtest for a couple of hours, and it didn't return any errors. i'll probably run it overnight, but i need my PC today. i thought i should try adding more voltage to it, but why should it need more voltage at lowest settings? maybe it is bad?

fouldsy 11-02-2005 02:03 PM

Try a memtest overnight for 10-12 hours to let it have a good few passes through each check. If the memory is both the same brand and speed, doesn't seem like a configuration issue, but more likely a bad memory module. Have you tried removing your original ram stick and only leaving the new one installed and seeing if the same problems persist?

slackhack 11-02-2005 02:17 PM

Quote:

Originally posted by fouldsy
Try a memtest overnight for 10-12 hours to let it have a good few passes through each check. If the memory is both the same brand and speed, doesn't seem like a configuration issue, but more likely a bad memory module. Have you tried removing your original ram stick and only leaving the new one installed and seeing if the same problems persist?
haven't done that yet b/c i needed to do some stuff today and didn't have time to experiment with taking modules in and out, etc. right now i'm compiling a new kernel with highmem support, as i noticed my old kernel config didn't have that enabled. that might account for the "missing" memory, at least. then i'll let memtest run overnight with just the 512 stick, see what that shows. thanks.

Back_to_Linux 11-02-2005 02:25 PM

hello,
are you try to use a 512 bank only ?
May be your mother board need to adress the same range of adresses, at the twice 256, acces isn't the same vs 512 'sticks'.
May be the twice consumns more. or, if it's the same mark, the technologie isn't the same !!

slackhack 11-02-2005 02:51 PM

Quote:

Originally posted by Back_to_Linux
hello,
are you try to use a 512 bank only ?
May be your mother board need to adress the same range of adresses, at the twice 256, acces isn't the same vs 512 'sticks'.
May be the twice consumns more. or, if it's the same mark, the technologie isn't the same !!

what does that mean, address the same range of addresses? is there a bios setting i can tweak to manipulate that? or maybe you mean i should run 2x512 instead of 512 + 2x256?

it's possible the technology isn't the same. i had never really heard of buffalo tech, but someone recommended it to me a couple of years ago, when i got the 2x256. it's been great up till now, it could OC 200FSB at turbo, and a little more with added voltage. maybe the newer stuff is from a different factory, different process, etc. all are labeled PC3200 cas 2.5, however.

recompiled a kernel, 2.6.14, still 0 highmem.
Code:

[2] sero:~ $ dmesg |grep mem
Memory: 906072k/917504k available (2060k kernel code, 10980k reserved, 569k data, 160k init, 0k highmem)
PInS memtype = 5
Freeing unused kernel memory: 160k freed


slackhack 11-03-2005 03:12 PM

okay, things are running better now with more voltage and slightly less aggressive timings. the nf2 board doesn't like cas3, so that was the problem there. so aside from that, i guess mostly it was just a matter of needing more voltage with the extra 512. merci! :cool:

Back_to_Linux 11-03-2005 03:30 PM

you need to see the serigraphie on the chip to know if they made by the same factory.
I know that for a factory, the chip's adressability vary. If the system are 3 chips to adress in a bank, it isn't the same algorythm to adress 6 chip in 2 bank, rigth ? your mother board and her bios don't communicate the same maner to the 512' bank, than they with the too 256 bank. It's what i meand. Even, we have the simple mind to aks uselves this the same, the system create a map of the memory to acces it. Phisycaly, this map reside in the chip, you understand what i mean? Try, like a precedent post says you, to use you 512 memory stick alone in your box make some "hard" test to know if all the chip in the stick are valid for first, and to know, for second, the maner the chip is build (true rate, clear voltage, number of chipset, vendor, and so on). make the same thing with your precedents stick alone, and compare their compatibility. An evolution in a chip, even with the same factory, may be represent some architecture different, because the assembler have a custom chip's programmation.

slackhack 11-03-2005 04:29 PM

thanks, BtL. it's still reporting as under a gig, does anyone know why that might be? :confused:

Code:

[2] sero:~ $ free -om
            total      used      free    shared    buffers    cached
Mem:          885        199        685          0        10        102
Swap:          964          0        964



All times are GMT -5. The time now is 05:22 AM.