Observations after using several memory diagnostic tools: Don't rely on one!

Over the past two years, I've had quite a bit of experience with diagnosing faulty memory modules--some very faulty, and some only marginally bad. One thing that I've noticed is that Memtest86+ is not always the best diagnostic tool. I know some of you will say that this is common sense, but the fact of the matter is that when I see people on these forums who suspect faulty memory, the usual suggestion is to run Memtest86+ to rule it out as the culprit. There is this line of thinking among many that Memtest86+ will show errors when memory is bad, and that if Memtest passes while Prime95 fails, the CPU must be unstable. Well that is simply not the case.

The main point I want to make is that some faulty memory modules can be a pain to find, and using only one memory diagnostic tool can give misleading results, and will only cause more delay and frustration in finding the problem. I have RMA'd 3 sets of Ballistix modules over the past couple of years, and only once was Memtest86 able to find the bad module of the two. For the other 2 sets, I used Prime95 torture test (passed with one module, failed with the other). In these cases, Memtest86+ could loop for 24+ hours without showing an error.

Most recently, I discovered another very useful memory diagnostic tool: Windows Memory Diagnostic. This handy utility is included on the Ultimate Boot CD (a must-have for any PC enthusiast!). In the case of my most recent memory failure, I found that Memtest86+ would run 12+ hours showing no errors on a faulty module; however, Windows Memory Diagnostic showed errors almost instantly!

So, the next time you are trying to diagnose an instability problem and you suspect faulty memory, don't be satisfied once you've passed only one memory diagnostic test! These tests are not as robust as one might think.

KillrBuckeye said:

Over the past two years, I've had quite a bit of experience with diagnosing faulty memory modules--some very faulty, and some only marginally bad. One thing that I've noticed is that Memtest86+ is not always the best diagnostic tool. I know some of you will say that this is common sense, but the fact of the matter is that when I see people on these forums who suspect faulty memory, the usual suggestion is to run Memtest86+ to rule it out as the culprit. There is this line of thinking among many that Memtest86+ will show errors when memory is bad, and that if Memtest passes while Prime95 fails, the CPU must be unstable. Well that is simply not the case.

The main point I want to make is that some faulty memory modules can be a pain to find, and using only one memory diagnostic tool can give misleading results, and will only cause more delay and frustration in finding the problem. I have RMA'd 3 sets of Ballistix modules over the past couple of years, and only once was Memtest86 able to find the bad module of the two. For the other 2 sets, I used Prime95 torture test (passed with one module, failed with the other). In these cases, Memtest86+ could loop for 24+ hours without showing an error.

Most recently, I discovered another very useful memory diagnostic tool: Windows Memory Diagnostic. This handy utility is included on the Ultimate Boot CD (a must-have for any PC enthusiast!). In the case of my most recent memory failure, I found that Memtest86+ would run 12+ hours showing no errors on a faulty module; however, Windows Memory Diagnostic showed errors almost instantly!

So, the next time you are trying to diagnose an instability problem and you suspect faulty memory, don't be satisfied once you've passed only one memory diagnostic test! These tests are not as robust as one might think.

Click to expand...

KB,

As you well know me from the Asrock forum, what you have stated is "spot on." I've had my share of checking RAMs that I thought were faulty. Some were and some were due to my own ignorance of voltage and timings requirements.

What I have been taught as it was explained to me is Memtest86 is a good start. If it errors there, don't bother going any further. I've had numerous instances exactly as you stated, they would run in Memtest86 forever and fail either in SuperPi or Prime many times.

If one is tweaking memory and timings, and SuperPi or Prime fail, it's not necessarily bad memory but incorrect timings, voltage issues or vcore issues. This is where the headaches begin to narrow down what the problem is.

Anyhow, this has worked for me, used in this order if timings, vdimm, vcore are correct and corresponds to what you stated:

1. Memtest86 from a boot disk
2. SuperPi (32 mb only)
3. Prime95

With new RAMs you need to know if they are good at default before anything else. If you pass all three, there is an excellent chance you have good RAMs and can start your overclocking. I have later passed SuperPi only to fail in Prime95 so it's a gradual series of good tests.

I won't go into how many Memtest86 passes one should do. I do 4 or 5 full passes to get started at default RAM and CPU speed. To me it doesn't make sense to OC right away. You need to know if RAMs are good at advertised default speed. If they aren't, there is no sense trying to overclock them.

Aruguably, some will dispute only using SuperPi 32 mb only. I want the best stress possible to be sure so I only use SuperPi 32 mb full cycle. If they pass at the 1 mb test but fail at 16 or 32 mb, what have you accomplished? There is still a problem. This is only my method and my opinion.

Prime95 is another arguable length of time. Usually a failure will show up within the first 10 - 15 minutes, other times I have had them appear after an hour or two. Keep in mind that many times when tweaking you may be right on the "edge" of a timing or voltage issue and it may take some time for the problem to appear. When I am satisfied I am where I want to be then I'll run Prime95 for 6 - 10 hours and I'm set to go.

Many others more experienced than me will have other methods that work. The above has worked for me and I couldn't agree with you more:

A series of good tests will enhance the chances of having good RAMs that you can depend on. It takes more than one to be sure and Memtest86 alone just won't prove anything but it's where to start. Unless you like to reinstall your operating system fresh or from a backup due to corruption, pick a few others to run in Windows such as SuperPi and Prime95 after you pass Memtest86.

There is a wealth of information here on this subject from those who have learned the hard way over the years. I use their experience all the time. You and others have guided me well and I always follow good advice of those more expereinced than me! RT

bump

MT86+ might not be the best absolute memory tester but it's a great diagnostic tool for telling you what needs tweaking in terms of timings, and voltages needed for memtest stability, you then will move onto windows stability tests. My opinion http://i4memory.com/showthread.php?t=6

eva2000 said:

MT86+ might not be the best absolute memory tester but it's a great diagnostic tool for telling you what needs tweaking in terms of timings, and voltages needed for memtest stability, you then will move onto windows stability tests. My opinion http://i4memory.com/showthread.php?t=6

Click to expand...

I agree with you for the most part. It's just that there are many people will tell you that Memtest86+ is all you need to determine if your memory settings are stable, or if you have a faulty module. In another thread some people were claiming that if Prime95/Orthos errors out, but Memtest86+ loops indefinitely without an error, then the memory definitely is not at fault. In my experience this is certainly not the case.

KillrBuckeye said:

I agree with you for the most part. It's just that there are many people will tell you that Memtest86+ is all you need to determine if your memory settings are stable, or if you have a faulty module. In another thread some people were claiming that if Prime95/Orthos errors out, but Memtest86+ loops indefinitely without an error, then the memory definitely is not at fault. In my experience this is certainly not the case.

Click to expand...

KB,

To add to your post, these three items are important:

1. Be patient
2. Be methodical
3. Write it down!