bad memory?

[gid]

My linux machine used to be pretty solid and stable... I've used this stick of ram for a few maybe 7 months I think. Recently, random things started crashing, mostly cron related stuff. mrtg, fetchmail, php, perl scripts. I can no longer complete a kernel compile without error and also large java programs are suddenly very prone to crashing.

 

So I ran memtest86 3.1a and it came back with 14 errors (I think that's the number, but am unsure, I can re-run if need be)

MB: AOpen AK77Pro

CPU: Athlon 2600+

RAM: Corsair VS512MB400C3

 

Once I swapped out memory with a known good stick of memory (also corsair memory, CMX512-3200C2) everything runs great without crashing. (so far at least)

 

Advice? Should I RMA?

 

EDIT bought from newegg

EDIT2, always running stock speeds, no overclocking

[RAM GUY]

Please make sure that you have the latest BIOS for your MB and load Setup/Optimized default settings, and try the following BIOS settings;

VS1GBKIT400C3 or VS512MB400C3

CPU Freq: 166 MHz (333 FSB)

System Performance: User Define

Memory Frequency: 100%

Resulting Frequency: 166 MHz (333 FSB)

Memory Timing: User Define

SDRAM CAS Latency: 3T

SDRAM RAS to CAS Delay (tRCD): 3T

SDRAM Row Precharge (tRP): 3T

SDRAM Active to Precharge Delay (tRAS): 8T

DDR Reference Voltage: 2.7 Volts

AGP Voltage: Default *unless you have ATI (9200 or 9600) or NVIDIA (5200 or 5700) then 1.6 Volts suggested!*

All other settings are motherboards defaults

Then please test them one at a time with http://www.memtest.org and let’s make sure it's not some other issue! I would run the test for at least 2-3 passes to be sure!

[gid]

I have the latest bios for my MB, but a lot of the DRAM timings settings don't match up. Leaving at stock voltage since I have a GF3. I actually have a 133 FSB Athlon XP 2600+ (older cored one)

 

Here are my DRAM timing options:

 

DRAM Clock

PCI CLK x3

PCI CLK x4 (default)

By SPD

 

SDRAM CAS LATENCY TIME

2.5 (default)

2.0

Auto

 

DRAM Queue Depth

2 Level

4 level (default)

3 level

 

DRAM Command Rate

2T command (default)

1T command

 

 

I downloaded memtext x86+ 1.40 and ran it against the swapped out (good) stick first (the Corsair XMS) since that's what I already had in the machine because it's stable. Ran for 2 passes, got 0 errors.

 

Next I put the Corsair VS memory back in, ran for over 2 passes, got 1117 errors on Test 5, and 10 errors on Test 8.

 

With my weird DRAM timing options which ones should I select?

[RAM GUY]

Please try these settings.

DRAM Clock

By SPD

 

SDRAM CAS LATENCY TIME

Auto

 

DRAM Queue Depth

4 level (default)

l

 

DRAM Command Rate

2T command (default)

And if you have an option to set the Dim Voltage, please set it too 2.7X Volts.

[gid]

I set my bios to settings you suggested, after about 2.5 passes of memtestx86+ 1.40 I got 3628 errors on test 5 and test 8 gave 14 errors.

 

Interesting that all of the errors seems to be in the 500MB plus range. Which would make sense. Linux itself rarely crashes (although yesterday morning it was locked up solid), because it resides in the lower good memory, but other user programs and what not are more prone to crashing because they have a greater chance of using the bad memory due to the program using up lots of ram (java), or the program being run many times (crontab entries). Of course this is just a theory, I've no idea what algorithm linux uses to decide what memory to use for what, but lowest address range first makes sense to me.

 

Any other ideas/work arounds?

[Wired]

Sounds like you haven't tested with ONLY the "bad" stick in yet. Try that first.

[gid]

I'm only ever using one stick of ram in the machine at a time. When I put in the single stick of value select 512meg, it fails memtest and programs crash.

 

The XMS ram that I pulled from another machine runs fine without error in the same machine and passes memtest.

[RAM GUY]

Are you trying to mix the XMS module with the Value Select? If this is the case, then all I can tell you is that it may or may not work, but we can not guarantee it. Now if the VS module is producing error when it is installed by itself, then I would suggest we get it replace. Please follow the link in my signature “I think I have a bad part!” and we will be happy to replace them or it!

[gid]

Ok, I RMA'd the VS ram and just received it today. It's still giving me errors on memtest, and boot fails. So I took out the VS ram and put the XMS ram back in the machine, memtest passes and boots and runs fine again.

 

Is it possible I got a bum stick of RAM returned to me? or do you think something else is going on here? CAS latency shouldn't matter, should it? The XMS ram is C2, while the VS is C3. My motherboard looks like it only has the option for SDRAM CAS Latency Time of 2.5, 2, and auto. Everyone I ask says this shouldn't matter, but I'm starting to wonder....

 

I'm thinking it CAN'T be that the VS C3 ram is incompatible, because this machine ran great with it for 6 months or so.

[RAM GUY]

Please try and clear the bios with just the vlaue select installed and then load setup defaults and exit saving chnages and test it with http://www.memtest.org to be sure! If you can I would maybe try and test the module in another system as well.

[gid]

Please try and clear the bios with just the vlaue select installed and then load setup defaults and exit saving chnages and test it with http://www.memtest.org to be sure! If you can I would maybe try and test the module in another system as well.

 

I tried clearing the bios settings again, the memory module still returned errors.

 

Then I tried it in a different machine and it works fine in the new machine. But unfortunately, when the VS is paired with another stick of 512 (XMS) in the new machine, it only runs at DDR333, not DDR400 like my two sticks of XMS run, or like the XMS by itself runs.

 

You think the motherboard from my orig machine just got picky all the sudden? I tried reseating the ram numerous times.