View Full Version : POS

10-02-2003, 03:22 PM
Whenever I install Win98 on my computer, it runs fine, apart from the occasional crash (not just a normal blue-screen). Though when I try and install XP, it crashes sometimes in the install, and (if I get far enough) it crashes randomly on the desktop. This is after installing both (98/XP) on two different drives, partitioned in a variety of ways - this makes no difference. Different HardDrives, and video cards and modems have been tested- none of which has helped. A virus check has also been performed. The memory has been checked too. The computer temperature seems alright. This has occurred in a different case and power supply too. What could it be? I remeber installing XP a while ago and it went alright.

The specs are:

Abit VA6 mobo
PIII Coppermine 667
some generic TNT2 M64/32mb
2 seagate HD 20gb/40gb

10-02-2003, 03:26 PM
It could be the CD - Try giving it a good clean and installing again?!

Graham L
10-02-2003, 03:52 PM
Sure, clean the CD. But if that still doesn't work, I would strongly supect your memory. Memory testing is a very difficult thing to do. The most severe test it usually gets is the installation of an OS. :D

10-02-2003, 03:57 PM
Why is that?

Graham L
11-02-2003, 03:00 PM
That is because installing the OS involves decompressing lots of BIG files. Big files means that lots of memory is used, so most of it will be used.

Decompressing requires that what in memory is exactly correct. The decompressor will complain or fall over if even one bit is wrong. The OS might give an error message if it can't read a block of data from a disk. The error testing on that is minimal. Many types of errors can "selfcancel" and fool a checksum or CRC check.

An even better memory test is to compile the Linux kernel. There's an FAQ about it in the LDP.

11-02-2003, 03:11 PM
There is a program called DocMemory (I think it is called) which does a pretty thorough test while booting from floppy.

11-02-2003, 03:12 PM
So that could be why installing XP seems to fail on an old PC that has had its case opened up quite a fair bit.

Would you suggest I use some other RAM to test out during install and see if that helps then?

Graham L
11-02-2003, 04:53 PM
Memory testing programmes prove that you can run memory testing programmes using the memory. They don't generally prove that the memory is error free; they prove that the memory passed the tests they set.

"Absence of evidence is not evidence of absence". Many years ago I ran a memory tester for hours on an 8k word core memory board which had been repaired (after waiting months for slice amplifiers). Not a niggle. Unfortunately, when I loaded the programme which was the sole task of the computer, it wouldn't work. I concluded that if we wanted to run memory tests, we could have that card in. If we wanted the computer to do useful work, we had to have that card out.

About the same time, HP were starting to use 4kx1 bit dynamic memory chips in their minis. They did a lot of research, and finished up using 21 bits wide for a 16 bit memory, and had ECC circuitry. All memory errors were logged, and part of the routine maintenance was to check those logs, and replace any chips which had had errors (even correctable). All the chips were socketed. The idea was that the memory boards would get better over time, as the "weak" chips were weeded out. They found little things like pattern sensitivity, which made "memory tests" useless, because they would take impossibly long times to go through all possible patterns of bits. Now, with multimegabit chips, the problem must be worse.

Most of the time if the "low" memory (where OS and application programmes mostly live) is OK, and there are a few flakey locations higher up, you will get away with it. If you are doing wordprocessing, and a few characters are wrong, "you know you can't spell" :D. If you have a spreadsheet, and a few of the numbers are wrong, well, "they're only numbers" :D.

But "upgrade" to XP, and the OS crashes often, that could be because essential system code is being clouted. There's more code, so it uses more memory. That's if you have succeeded in installing it at all, because of the decompression test.