Memtest question - should I be worried?

[Drizzy]

Background: I had 1 GB of ram (2X512) in slots 1 and 3 running on my DFI Ultra-D mobo. This ram is Mushkin PC3500. I recently purchased another GB of ram (2X512) and put it in slots 2 and 4. This ram is PQI PC3200. Then I ram Memtest right after installing the second GB of ram. Here are my results:

Ran the test 15 times and 13 passed with 2 failed. The 2 that failed were at the same spot and same error.

0006fdd6fc4 - 1789.3MB and the error was on test 8 with pattern ffffffff -> fffffffef.

Here are my questions -
1. based on the location of the error would I assume its in the last slot because its the last portion of my 2 gb?
2. Is it normal to get an error with this much ram and a 7-1 pass rate or is ANY error too much?
3. Should I run my ram in a different configuration like the same type of ram together - slot 1 and 2, then slot 3 and 4?

Any help is appreciated - thanks!

 

[John]

Any error is a bad error. Try the new memory by itself, both in different slots, raise the voltage a notch, take off SPD and set manual relaxed timings, etc.

 

[eagle101]

Originally posted by: John
Any error is a bad error. Try the new memory by itself, both in different slots, raise the voltage a notch, take off SPD and set manual relaxed timings, etc.

john:can you guide me to the memory test to download......i hope i dont get errors thanks!

 

[Seekermeister]

http://www.memtest86.com/

 

[Minerva]

Repeatability in memory error results usually indicates weak spots. Raising the voltage and/or relaxing timings may work. Random errors usually indicate running too fast (out of spec) or voltage too low for current speed, etc.

Raise the mem voltage and run again. If error free, let it run like that but check it a few times over the period of a few weeks to see if errors come back.

I've seen pc's that give one or two errors run windows perfectly fine for years and then I've seen ones that BSOD a few times a day when tested produce THOUSANDS of errors nearly immediately when the test starts. Problem is one day you could be working and that failing address is called and POOF byebye work.

 

[Lemon law]

Its also my understanding that some mem test errors on certain tests result from various processor mobo combinations---and the same ram shifted to another processor mobo combinations might well be error free---and the memtest86 homepage somewhat states that if the ram passes the opening screen initial test after a number of runs---its pretty well
error free for the most part.---but if you get many errors after a run of the initial memtest86---you can pretty well say your ram is junk.---but it will take some ram module shifting to determine which module it is assuming you have more than one stick.

And the other unmentioned factor---is your ram ECC or non-ECC?

 

[John]

Originally posted by: Seekermeister
http://www.memtest86.com/

That's the old version that the new and improved Memtest86+ is built upon.

-= History =-

The first version of Memtest86+ was released on early 2004, based on memtest86 v3.0 that was not updated since mid-2002. Our main challenge was to provide an up-to-date version of this useful tool, as reliable than the original. Our work started when we got the first AMD64 system. Unfortunatly, the original memtest v3.0 didn't run at all. After looking at the source code, we fixed the bug.After some days, I saw lot of other things like chipsets or CPU that were not correctly detected or not detected at all. As I'm the chief-editor of a french hardware website (www.x86-secret.com), I have access to lot of recent hardware and I can test and debug on quite all available motherboards on the market. After adding detection for all current CPUs, I've added detection for all current chipsets (SiS, VIA, nVidia, Intel) and ECC Polling for AMD64, i875P and E7205. Then, I decided to display some useful settings for the most popular chipsets. For exemple, on i865PE/i875P series, memtest86+ will now display FSB & Memory frequency, PAT status, memory timings, ECC status and the number of memory channels. Next version will perhaps contain several enhancements and bug-fixes.

http://www.memtest.org/

 

[Drizzy]

Running non-ECC ram.. upped the voltage and that seemed to fix it. I'll run another set of tests overnight tonight then.