1 |
On Monday 12 February 2007, Matt Richards wrote: |
2 |
> I wouldn't of thought it would be the RAM not from behavior like that |
3 |
> but it does sound like a overheating issue, I have had computers just |
4 |
> power off because they get too hot. |
5 |
|
6 |
The thing with memtest is that it does (mostly) predictable tests, which |
7 |
can detect consistent errors with memory chips, or somewhat random |
8 |
errors with memory. |
9 |
|
10 |
BUT, it doesn't help at all with detecting didgy hardware in other |
11 |
areas. It doesn't give you any valid tests for what happens when the |
12 |
machine gets hot for example. Compiling a kernel (or openoffice.org) 10 |
13 |
times in a loop does do that. |
14 |
|
15 |
So, I would say that you have shown that your memory chips are not |
16 |
likely to be faulty, but you have not yet proven that the cpu (or the |
17 |
rest of the machine) isn't getting hot and overloading. |
18 |
|
19 |
Jeff, if you were paying me to diagnose your problem, I'd be telling you |
20 |
to swap out the psu and test some more, then disassemble the |
21 |
cpu/heatsink combination and replace the thermal grease and test some |
22 |
more. If it then fails, the tests are still not conclusive so I would |
23 |
want to forcefully cool the machine (with an external fan etc) while |
24 |
putting it through some stress tests (like a looooooong compile). |
25 |
|
26 |
Unfortunately, it happens to be true that only extended periods of high |
27 |
load using all (or most) of the systems in a machine qualify as a |
28 |
decent test. |
29 |
|
30 |
alan |
31 |
|
32 |
|
33 |
-- |
34 |
Optimists say the glass is half full, |
35 |
Pessimists say the glass is half empty, |
36 |
Developers say wtf is the glass twice as big as it needs to be? |
37 |
|
38 |
Alan McKinnon |
39 |
alan at linuxholdings dot co dot za |
40 |
+27 82, double three seven, one nine three five |
41 |
-- |
42 |
gentoo-user@g.o mailing list |