How to debug crashing system?

Y

you2

[H]ard|Gawd
2FA
Joined
Aug 10, 2004
Messages
1,423
I built a system a couple of months ago and it has been rather unstable (getting worse). The components are:
--
nf4 ultra-d
fsp 600w-gln (over kill)
x1800xt 512mb
2 1gb G.skill D500 F1-4000USU2-2GBHZ
san diego 3700+
1 optical eide
1 74gb raptor
--
I am not overclocking - memory is using auto defaults (but look correct from atool/cpuz) and proc is at 2.2 ghz. Clearly something is bad (memory, cpu, psu, mother board).

Any suggestions of how to find/figure out what is busted? I do not have a probe so i cannot measure voltage directly but i am running mbm5 which seems to indicate reasonable values. Again according to mbm 5 max cpu temp is 45 with mean around 42. max video temp is 81 (which is normal for 1800xt) north bridge is 50c.

I just want to figure out which component is bad and fix it !

Thanks,
You
 
how does it crash? is it like hard crash and you have to reset? or does it restart? Well.. I'd start from taking out any unnecessary hardware (NIC, sound card - whatever you have there..). Test memory with memtest86.. it might also be psu going bad - check the voltages on 5 and 12 line with a multimeter..
 
With prime95 after about 45 minutes it will detect an error. I've run memtest86 for two iterations without an error (not that long) but I haven't run it overnight yet. With games that stress the system (like civ4/quake4) will bsod (cause reboot). After the game crashes - crashes will become more frequent until the system is left off 30 or 45 minutes (this sort of suggest heat issue but the cpu never gets above 46 - northbridge 50). Obviously if prime95 detects and issue there is something wrong (probably one of memory/cpu/psu/motherboard). What make sit painful is that you have to run it a while to get an error which make diagnostics (to me) unobvious.
 
Run memtest for at least 6 hours then come back. I wouldn't be surprised if it's not getting enough voltage.
 
What is a safe voltage for that memory? the dfi board will allow me to up it. However, I rather not fry things...
 
Ok I ran memtest86 last night for 8 hours. No errors detected. Bleh. I almost wished it found an error.
 
Any suggestions at all? I don't mind trying to replace something - just not sure if i should start with the memory, cpu or motherboard (I hope the psu is ok).
 
One way to check if its a heat problem is to run it open case with a room fan blowing cold air into the case.

Try three or four different stress testing programs.
Do they all crash ?

How old is the Windoze install ?
When did you lass check for mal ware ?
Are ALL your drivers up to date and any old ones compleatly removed ?

Whats does the BSOD say.

Luck........... :D
 
Am I understanding correctly that the GPU is running 81 degrees C???? That sounds pretty hot to me. Even if the card can stand it, heat given off may be affecting nearby mobo components. That would fit with the increasing frequency unless you allow a long cooling period.

One thing you can do is CAREFULLY feel around on the motherboard to locate hotspots. How about HDD temps???

CoachB
 
The GPU peaks at 81C which is normal for ati. However, when I run prime95 or similar it does not get very hot. I've tried feeling around the motherboard - someone suggested it might be the ram but it did not get that hot - however, the cpu heatsink at base does feel quite warm. Perhaps a little more than the reported 43C.

The S.M.A.R.T thingy seems to indicate the hard disk is around 31 - not sure how accurate it is - i've begun to wonder if perhaps the swap disk is on a bad block. There is a 120 fan (li an case) that blows directly onto the hard disks.

Question - even though prime95 dies - if memtest86 runs for 8 hours without error can I assume it is not the memory that is bad and something else that is wrong - or could it still be the memory ?
 
Ok. I think the problem was ram. Annoying as heck that mem86 runs fine overnight. If I put in one stick or ram (stick a) the system dies on prime95 after bout 2 minutes. If I put stick in (stick b) it runs fine for 20 minutes. I'll run an overnight test to make sure this is the case and then repeat with stick a to make sure but I think this is the problem. (I was able to er run my favorite application on stick b for 2 hours without an issue and prior to pulling stick a it would run no more than 20 minutes.
 
You must log in or register to reply here.
Back
Top