| Forum Home | ||||
| Press F1 | ||||
| Thread ID: 36829 | 2003-08-21 09:58:00 | What price memory testers? I'm not so sure! | Billy T (70) | Press F1 |
| Post ID | Timestamp | Content | User | ||
| 169262 | 2003-08-21 09:58:00 | Hi Team Anybody who followed my suspected partition failure (pressf1.pcworld.co.nz) thread will now know that the actual cause was a defective 256MB Dick Smiths memory module. To find out what was wrong with the module I located and downloaded this memory tester (oca.microsoft.com) from Microsoft. Surprisingly, it found no fault at all despite several passes, but then it doesn't give any indication of the memory size, at least not with the standard test so it couldn't tell me it was now only 128MB. Even more surprisingly, a much better indication was achieved in this particular instance by enabling the boot sequence RAM check. This beeped and chirrupped, then declared a total of only 128MB. The other module showed 256MB without the dawn chorus. The faulty module was in Bank 0 of an Intel SR440BX motherboard and swapping it over to Bank 1 let the boot check count the full available 384MB. However, it still beeped like crazy at the end of the RAM check so I decided not to try my luck by letting it go any further and try starting Windows. Seems strange to me that a dedicated RAM checking program didn't give any indication that the module was faulty, but a boot RAM check picked it straight away. Is there any rational or logical reason for that peculiarity? Cheers Billy 8-{) |
Billy T (70) | ||
| 169263 | 2003-08-21 10:06:00 | Probably because once Windows got loaded up the computer was kindly telling Windows that the RAM was 128mb and ignoring the fact that it was actually a 256mb, so the RAM tester was only testing what was being made available to it and Windows by the PC. Mike. |
Mike (15) | ||
| 169264 | 2003-08-21 10:54:00 | It could be how the memory is managed, part of the memory is stored away for the OS, it could contain the faulty part, it's not wise to have a program that could probe the OS ram and accidently throw it into a crash course. Well that's how I'm seeing why the program never worked... but why would I stick up for MS, it probably never worked in the first place :P, make it open source and you'll see why it never worked. |
Kame (312) | ||
| 169265 | 2003-08-21 11:11:00 | Sorry guys, I should have made it clear that it is a RAM tester that operates off a boot disk . I thought that point was covered by my comment that I didn't allow it to boot to Windows, however there are none so blind as those who wrote the post . :8} Cheers Billy 8-{) :| |
Billy T (70) | ||
| 169266 | 2003-08-21 12:06:00 | The memory testing program that I suggested is Memtest86 version 3 . 0 and it is the one that the Overclockers forum use . After your results with the MS one I am not sure whether you would want to try it though . In case you do you can grab it from here (http://www . memtest86 . com/) . :-) |
Susan B (19) | ||
| 169267 | 2003-08-22 02:42:00 | I won't say "I told you so" now you say it was memory all the time . I'll just think it . :D Memory testing has always been diabolically difficult . I've struck this with core memory too . . . the board ran the memory diagnostic programme for hours without a problem . It just wouldn't work with real programmes . :_| I've always worked since on the principle that the best test for memory is running programmes which won't tolerate errors . That includes installing OSs, because that involves uncompressing huge files (decompressors are fussy), or the very good one of compiling the Linux kernel . B-) HP in the 70s had a great article about the memory for one of their new minis in a magazine they produced at the time . They were using the new 4k x 1 bit memory chips, rather than core . They finished up with using 21 bits for 16 bit words, and ECC which would detect and fix single bit errors and detect more than single bit errors . They also logged all memory errors, so the engineeer in his monthly visits could replace any (socketed) chips which had higher than normal error rates . The 4k x 1 chips had various causes for errors . . . apart from random cosmic ray strikes . . . including a particularly nasty mode: pattern-dependence -- a bit would give an error only if there was a particular bit pattern in the whole 4k on the chip . Since the number of possible patterns in 4096 is very large, it's impossible to test in time less than several orders of magnitude greater than the age (opr even the predicted lifetime of the universe . With gigabits on a single chip, it must be worse now :D . |
Graham L (2) | ||
| 169268 | 2003-08-22 09:09:00 | > Sorry guys, I should have made it clear that it is a > RAM tester that operates off a boot disk. > I thought that point was covered by my comment that I > didn't allow it to boot to Windows Well all that said, my point still stands - once past the POST the computer tells everything there is 128mb, so the tester would test the RAM that is available to it. Mike. |
Mike (15) | ||
| 169269 | 2003-08-22 09:55:00 | Yes, I understand the point you were making now Mike. I guess that means you can't fully test any RAM unless you have some sort of external test rig. Cheers Billy 8-{) |
Billy T (70) | ||
| 169270 | 2003-08-22 09:58:00 | > I guess that means you can't fully test any RAM unless you have some > sort of external test rig. I would assume so... but I'm probably wrong on all counts :) |
Mike (15) | ||
| 169271 | 2003-08-22 09:59:00 | Mike. | Mike (15) | ||
| 1 | |||||