SDRAM problem

[fantrace]

I can't quite fathom out what is going on here. I'm rebuilding an old system based on a Gigabyte BX2000+ MB with 650MHz P3 running at 6.5*100MHz. Some of the original RAM was definitely faulty so trying to add a couple of 128M PC133 SDRAMs. I've also still got a couple of old 64M modules and a 32M one. With the 2 new 128M plus the 2 old 64M modules installed, WinXP wouldn't install with error message consistent with faulty memory modules, so I've been running the Windows Memory Diagnostic tool to try and work out what's wrong.

All the memory passes basic/standard tests no problem, but the new 128M sticks fail on one or other of the two extended tests that test the memory with cache off (MATS+ and Stride).

The first of 128M modules seems pretty clear cut - hundreds of fails, with addresses that end BB28, BB30 and BB38, all with a result of FFFFFF7F when FFFFFFFF was expected. Might have expected to see failures at addresses ending BB20 perhaps.

The second 128M module passes the tests when it is on its own, but if you add one or more of the 64M or 32M as well, then it fails consistently at 3 or 4 addresses - such as 06AA58E8, 060B58E8, 060E58E8 and 060F58E8, returning 00000040 when 0 was expected. I've tried all sorts of combinations of the old 64M/32M memory with this 128M module in the 4 MB sockets, and the addresses might shift, but the 3 or 4 failures always occur, returning 40 instead of 0. Again why sometimes only 3 failures when you might expect 4.

The old memory only in the various combinations seems to work fine.

I think the new memory must be faulty, but why does it pass the standard tests, then fail these two of the extended tests which have cache off? Or could it be a MB problem or simple incomatibility? It's suspicious that two brand new modules fail in such similar and slightly odd ways. I don't want to take it back and have the shop test it only to say it is fine, or for them to exchange it and end up with similar problems with the replacements.

Unfortunately my main machine doesn't use 168pin SDRAM, so I can't test the modules in a different machine.

So any suggestions on what might be going on please? Could it be timing related - is the anything in the BIOS (AWARD) that I could try tweaking to see if it helps? Should I just get the memory replaced and see what happens?

 

[Flinx]

You need to test the ram in another motherboard where everything is good. I've had a motherboard where one of the slots was no good! (partially burned coil or faulty capactor?)

Why are you rebuilding?

And have you looked at <A HREF="http://www.memtest.org/" target="_new">memtest86 ?????</A>

The loving are the daring!

 

[Crashman]

Several problems:

First of all, the BX is only compatable with low-density RAM. For 128MB, that means you need 8 or 16 chips per module, any less would be high density.

Assuming that you don't have a density problem, the BX may be too good a chipset for your RAM. It had very fast memory access times, and most newer modules are made for slower chipsets like the 810 and Apollo Pro 133.

You may be able to stabilize your system by manually selecting slower timings in BIOS.

Also, your board is quite old and may have one or more bad capacitors causing your memory bus to become unstable.

Finally, the problem could be caused by low voltage from your power supply, or fluctuating voltage on the 3.3v line.

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>