1 |
Andrey Vul schrieb: |
2 |
> On Fri, Nov 28, 2008 at 07:29, Florian Philipp |
3 |
> <lists@f_philipp.fastmail.net> wrote: |
4 |
|
5 |
>>> |
6 |
>> Iain's diagnosis seems about right. I would start with a nice 2h memtest86+ |
7 |
>> test followed with a 1h cpuburn test. |
8 |
> I believe you mean 2d memtest86+ unless computer has DDR9-99999 RAM |
9 |
> where 100 (full) cycles could be completed in under 2 hours. |
10 |
> Also, do cpuburn for a day or more - bad cooling could take hours to |
11 |
> mnifest themselves. |
12 |
> |
13 |
|
14 |
You are right. To be absolutely sure, you should take the time for a |
15 |
very long test. I've heard of errors which were first seem after a week |
16 |
of testing. But here we are talking about a problem which was visible |
17 |
after just one kernel compilation. I would bet that you could see it |
18 |
after just one test cycle of memtest and half an hour of CPUburn. |
19 |
|
20 |
One the other hand, I once had a memory error (due to too short |
21 |
latencies) which couldn't be found my memtest within a day but crashed |
22 |
Elder Scrolls Oblivion after just 5min. You can never be sure with those |
23 |
things. Who knows? Maybe the power supply is too weak and you need to |
24 |
stress RAM, CPU and hard disk at the same time (as a compilation would do). |