segfault due to underpowered (i.e. psu) system?

[Crazor]

Hi!
I've got a really weird problem over here. I'm trying to install Gentoo on 6 somewhat cheap AMD Duron 1,0GHz systems. Cheap means crappy boards w/ 3 PCI slots and no brand on them. Cheap does NOT mean crappy RAM, just to make sure anyone bothers asking me 'bout...

All six machines are identical to the bone. On every single one I get random sig11 errors during compilation. First guess: heat. No way, heat sinks won't even get warmer than the average human hand... Second guess: RAM. No way, memtest86 says ok (which doesn't really mean that much), but the RAM actually works fine in another system. I tried everything including BIOS update. By the way, the errors are not only random, but somewhat seldom as well, sometimes I can successfully build 3 Kernels in a row, another time not even a half.

I wonder if those elCheapo machines could suffer from underpowered PSU's, only just a bit too weak for handling high loads over time (AMD's just eat much power ). But everywhere I read about complete lockups caused by weak PSU's, so I ask all the gurus out there: May it possibly be the PSU?
I'll give out some specs tomorrow, when I get back to compilation hell . Just wanted to ask, cause this won't let me sleep *g*

[Schmolch]

All my segfaults - and i think i had lots - where caused by bad ram.
Bad CPU (due to overclocking) caused rather freezes and weak PSU rather hard-resets.
Bad ram means the ram-settings in the BIOS were set to "normal", which is slowest, but it still couldnt do that right though it was the most expensive ram from the shop (samsung only).

What did it for me was to change ALL ram-settings in the BIOS to the lowest and slowest settings possible.
After a while of experimenting i discovered which settings i had to lower specifically and which i could raise again.

[Jimboberella]

Segfaults in my rig are due to heat. I have solved that by adding a larger fan to the cpu. But bad ram or agressive ram timings can also lead to this.

[hielvc]

I ave simular probs with a el chepo MSI 6378 mobo. Like you said memtest86, prime 97 & setia all run fine, but compiling is real iffy. I was thinking about yours and my prob and how if I hit the floor and change my fsb to 100 ( I have an 1800-xp ) it builds fine at night. But Ive had large world updates that biuld fine at night to at fsb 133 and it dawned on me that it may not be weak power supplies but poorly filtered. I'm in biomed and the machines I work on count and size voltage pulses, be it thru the the appetures, radieted AM or line noise. Iv'e cant tell you how many times the service people have been around the block power because ofpower problems. It just occured to me that at night the power delivered to you and me is a lot cleaner, fewer microwaves AC, lighting loads swthing on motors kicking in and that might be why I can do large world updates, gnome xfree glibic gcc, and etc at night when I go to bed. but cant recompile my dandg kernel or anything during the somedays and other days I can. So the moral of this long a*s post is try at night :^) also try setting your boxs to build packages. Then you can gather all the packages onto one machine Oops another subject another day.

hielvc

[Crazor]

thanks for your replies..
so, to sum up, you guess either heat or ram are causing the problem...
i would say heat is not the problem, cpus don't get any warmer than probably 40°C, so this is really not that hot =)

ram.... umm... 6 computers, all identically fitted with components... so you really mean _6_ bad ramsticks in a row? i can hardly believe that. but i'll try swapping them with apparantly working ram.

i remember being told by some users that the computers sometimes completely fail when running windows 2k (i.e. screen turns black, then normal reboot). i'd say win2k was rather stable for all my setups using it, but somehow i didn't realize the story before
so when you say lockups/hard resets due to faulty cpu/weak power, could it be that somehow linux just survives but windoze didn't?

later i'll check what psu's are in the 'puters. i won't be surprised finding 120W or something that low in there =)

[Crazor]

update: 300W should be enough for 1,0GHz Duron w/ only the board, one hd and cd each. no other cards in the boxes.
changed ram to a working one and compiling still failed

seems to be the DFI AM33-EL board.
Any idea what i can try to get rid of the segfaults? may it be some ide setting in the bios causing the disk to randomly flip a bit of code?