unusual memory failure \ VOGONS

unusual memory failure

Topic actions

First post, by MatureTech

Posted on 2015-01-17, 13:31

MatureTech Offline

Rank Newbie

Rank: Newbie
Posts: 55
Joined: 2013-10-09, 13:51

I've seen bad DIMMs, bad memory slots, and otherwise good mobos that are just picky about memory, but this is a new one and I'm not sure what to blame as the root cause.

The mobo is KT133A. Has 3 DIMM slots supporting up to 512 MiB in each. I was running 1 GiB of mismatched PC133 memories, 512/256/256, at conservative settings. Recently acquired 4 matching 512 MiB CL2 DIMMs with both Crucial and Micron labels on them. Installed 3 of them and the problems began.

After testing lots of combinations I found that the trigger is not which DIMM I install or which slot I use, but simply whether I install more than one. One DIMM passes memtest with the most aggressive settings in BIOS. More than one DIMM fails even with the most conservative settings in BIOS. The failures are not total; MemTest86+ gets down to Test 7 if only two DIMMs are used, and down to Test 5 if three DIMMs are used.

Anybody else ever seen that? Lacking any new evidence, my best guess would be that the memory controller is made of suck and gets flaky when loaded down with 512 MiB modules. However, its marginal performance might in turn be blamed on a marginal mobo or power supply.

ISA go Bragh™

Reply 1 of 6, by AlphaWing

Posted on 2015-01-17, 14:36

AlphaWing Offline

Rank Oldbie

Rank: Oldbie
Posts: 703
Joined: 2014-06-02, 15:34
Location: USA

4.20 or 5.01?
4.20 can give false errors on some via chipsets.

Reply 2 of 6, by joacim

Posted on 2015-01-17, 15:33

joacim Offline

Rank Newbie

Rank: Newbie
Posts: 92
Joined: 2014-07-08, 00:55

Depending on which version of GCC Memtest86+ was built with, it can give false positives on test #7. The bootable binary from memtest.org should be good.

I changed RAM and CPU multiple times, so I just figured I had a bad motherboard. That was until I noticed the same errors in test #7 on multiple different computers.

Reply 3 of 6, by AlphaWing

Posted on 2015-01-17, 16:13

AlphaWing Offline

Rank Oldbie

Rank: Oldbie
Posts: 703
Joined: 2014-06-02, 15:34
Location: USA

Apparently 4.20 has issues with some modern intelchipsets too!
I've only seen it happen with my via boards.

Reply 4 of 6, by joacim

Posted on 2015-01-17, 16:41

joacim Offline

Rank Newbie

Rank: Newbie
Posts: 92
Joined: 2014-07-08, 00:55

Might be worth testing with a few different builds of memtest86+. Both from memtest.org and from other sources. Just to exclude differences in compilers and versions of memtest86+.

The errors OP got in test #7 might be the same false positives that I had. The errors in test #5 might be something else (via bug? 😀).

Reply 5 of 6, by AlphaWing

Posted on 2015-01-17, 17:19

AlphaWing Offline

Rank Oldbie

Rank: Oldbie
Posts: 703
Joined: 2014-06-02, 15:34
Location: USA

Test 7 errors are what I've encountered with it.
I use the floppy builds from memtest.org, most of the time.

Reply 6 of 6, by MatureTech

Posted on 2015-01-17, 18:18

MatureTech Offline

Rank Newbie

Rank: Newbie
Posts: 55
Joined: 2013-10-09, 13:51

It was a boot floppy for MemTest86+ 5.01. Test 7 was block move not random number sequence. I was getting single-bit errors. Still, this is interesting information and I will see what reproduces with other versions and with MemTest86 nonplussed.

Testing with two sticks of memory:

In SoftMenu III Setup, turned off all of the special tweaks, including "Force
4-Way Interleave" and "Enable Dram 4K-Page Mode."

In Advanced Chipset Features,
---DRAM Timing = Normal (the other conservative option was "SDRAM 8/10ns")
---DRAM Bank Interleave = Disabled
---Delay DRAM Read Latch = AUTO
---MD Driving Strength = Hi
---SDRAM Cycle Length = 3 (The DIMMs are rated CL2)
---Fast R-W Turn Around = Disabled

Memtest86+ 5.01 boot floppy says:
---322 MB/s (was 379 MB/s with aggressive BIOS settings)
---First failure is Test #7 [Block move] burst of single-bit errors below 32 MB.
Memtest86 4.3.4 boot floppy says:
---321 MB/s
---First failure is Test #6 [Block move] same error pattern.
Memtest86+ 4.20 boot floppy says:
---322 MB/s
---Test #5 [Block move] same error pattern.
Memtest86 4.3.7 booted from a CD says:
---same as 4.3.4.

Swapped the positions of the DIMMs.
Memtest86+ 5.01 boot floppy says:
---No change from before; the address of the errors is still below 32 MB.

ISA go Bragh™

Go to top of page Go to top of page

Back to General Old Hardware