Page 1 of 1

Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 1:09 pm
by churin
I am running Memtest86+ v.4.20 and noticed what are shown for the Settings are incorrect. The CPU die size as 65nm is also incorrect. The RAM in use is a pair of 4GB DDR3-1600 CAS 9-9-9-24. The screen shot of the Memetest86+ is as follows. What could be wrong?
Image

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 1:24 pm
by Glorious
Memtest86+ was last updated very early in 2011, the FX-6300 was launched very late 2012.

Simply put, it doesn't recognize your chip properly, because your chip is newer than it is.

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 1:25 pm
by bthylafh
That program is older than your CPU and chipset, and because of how it's written it doesn't know about hardware unless it's told.

This is newer and you should probably use it instead:
http://www.memtest86.com/

Yes, the version numbers are the same. That's for historical reasons to do with the MT86+ project forking MT86 when it went dormant several years ago.

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 3:02 pm
by churin
Thank you very much the replies. I will switch the utility to the one suggested.

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 3:23 pm
by just brew it!
Heh, I did not realize that PassMark had taken over maintenance of Memtest86.

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 4:17 pm
by churin
I am running PasMark Memtest86 v.4.30 now and seeing on it "Pass complete, no errors, press Esc to exit". Does this mean the RAM passed the test? If so why the utility continues running?
Image

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 4:22 pm
by bthylafh
Because RAM is a funny beast and some errors don't manifest unless you let the test run a long, long time. You're free to stop the tester, but you could also let it run overnight and be a bit more sure that your RAM is good.

OTOH I've had a few times when RAM was definitely bad but the tester didn't pick up on it, but the computer worked fine once the offending stick was replaced.

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 4:40 pm
by churin
bthylafh wrote:
Because RAM is a funny beast and some errors don't manifest unless you let the test run a long, long time. You're free to stop the tester, but you could also let it run overnight and be a bit more sure that your RAM is good.
Does the second pass and thereafter repeat the exact same test as the first pass?

OTOH I've had a few times when RAM was definitely bad but the tester didn't pick up on it, but the computer worked fine once the offending stick was replaced.
How did you determine the offending stick?

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 5:35 pm
by Captain Ned
churin wrote:
How did you determine the offending stick?

If you get an error, pull all but 1 stick. If it tests good, swap it for another. Lather, rinse, & repeat until you find the offender.

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 5:43 pm
by churin
Captain Ned wrote:
churin wrote:
How did you determine the offending stick?

If you get an error, pull all but 1 stick. If it tests good, swap it for another. Lather, rinse, & repeat until you find the offender.

I asked that question because he indicated that the tester failed to detect any error.

Re: Memtest86+ behaves strange

Posted: Wed Jul 10, 2013 6:16 pm
by Airmantharp
churin wrote:
Captain Ned wrote:
churin wrote:
How did you determine the offending stick?

If you get an error, pull all but 1 stick. If it tests good, swap it for another. Lather, rinse, & repeat until you find the offender.

I asked that question because he indicated that the tester failed to detect any error.


Still happens. I've been meaning to do it with my Corsair, just haven't gotten around to it yet; now that I've just done a rebuild, it's probably time :).

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 6:45 am
by bthylafh
churin wrote:
Captain Ned wrote:
churin wrote:
How did you determine the offending stick?

If you get an error, pull all but 1 stick. If it tests good, swap it for another. Lather, rinse, & repeat until you find the offender.

I asked that question because he indicated that the tester failed to detect any error.


It's as Ned said. Memtesters aren't flawless.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 8:38 am
by churin
Does every "Pass" repeat the exact same test?

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 8:40 am
by bthylafh
Natürlich.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 8:51 am
by churin
Then why do we need to let it repeat? Isn't once enough?

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 9:07 am
by nerdrage
Some of the tests are identical, some are changed randomly with each pass. Also, RAM failures tend to be intermittent. In order to find an intermittent problem, you need to run the same tests over and over again until the intermittent problem occurs.
RAM chips consist of a large array of tightly packed memory cells, one for each bit of data. The vast majority of the intermittent failures are a result of interaction between these memory cells. Often writing a memory cell can cause one of the adjacent cells to be written with the same data. An effective memory test attempts to test for this condition. Therefore, an ideal strategy for testing memory would be the following:

  • Write a cell with a zero.
  • Write all of the adjacent cells with a one, one or more times.
  • Check that the first cell still has a zero.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 9:21 am
by churin
nerdrage wrote:
Some of the tests are identical, some are changed randomly with each pass. Also, RAM failures tend to be intermittent. In order to find an intermittent problem, you need to run the same tests over and over again until the intermittent problem occurs.
RAM chips consist of a large array of tightly packed memory cells, one for each bit of data. The vast majority of the intermittent failures are a result of interaction between these memory cells. Often writing a memory cell can cause one of the adjacent cells to be written with the same data. An effective memory test attempts to test for this condition. Therefore, an ideal strategy for testing memory would be the following:

  • Write a cell with a zero.
  • Write all of the adjacent cells with a one, one or more times.
  • Check that the first cell still has a zero.

Thank you very much for enlightening me about RAM test.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 10:04 am
by JPForums
churin wrote:
Then why do we need to let it repeat? Isn't once enough?

The short answer is no.

While straight up defective chips should be detected with a single pass, chips with impurities that leave them functional, but effect performance have other considerations.
Transistors in memory chips (and all digital silicon chips for that matter) have rise and fall times that vary with environmental considerations such as voltage and heat.

Heat causes physical expansions in the silicon that, while microscopic, are significant when compared to the size of the transistor. If you have a chip that is already marginal, then even adding a small increase to the path length through a transistor can push it over the top. The actual amount of expansion is different between cases and obviously depends on your cooling solution, but it only takes one transistor missing timing to make your memory bad. So you must wait for them system to warm up before you can call it good.

The voltage supplied to the gate of the transistor also effects how quickly the transistor switches. Higher voltages turn on (or shut off) a transistor more quickly (hence its use in overclocking), but also increase power loss and heat dissipation. If a memory chip is marginal, a small droop in voltage at the wrong time can cause failures. Lower quality DC regulation in power supplies or in the voltage regulator modules on a motherboard can cause such droops. While these droops may occur with some low frequency, it is only when the transistor(s) in question are being accessed that such a droop will actually cause a failure. The length of time you must run the test to catch such an error depends on the frequency of droops and the probability that they occur when the marginal transistor is in use. The droops can also be cause by transients such as a large, sudden increase in power draw from events like spinning up a large array of HDDs all at once or suddenly going from a near power off state to full load with a (or several) high powered graphics card. A similar issue can crop up if voltages spike during access. Memory cells are tightly packed and a relative spike into one memory cell can bleed into surrounding cells potentially turning them on if they were off. Even an overclock on your processor can effect it. Though to be fair, power supplies and motherboards with good transient responses will go a long way to resolving such issues. These transient issues are often missed by dedicated memtesters as they require system loading not present in the memtester environment to cause the error. I agree with the overnight suggestion above as it is increasingly improbable that any issues not seen in that time frame will be seen by the dedicated memtester at all. Some of the 3DMarks benchmarks have shown themselves to be quite sensitive to overclocking issues, so that is another tool you can use, but they don't really use that much RAM.

Most memory errors where the chip isn't flat out defective (and thus generally easily detectible) are a combination of the above heating and voltage regulation issues on marginal chips. It follows that an increase in voltage, a decrease in timings, or a decrease in clock speed will often times resolve these marginal issues. Just keep in mind that you don't always have the ability to modify the part of the chip that is effected, though there are plenty of motherboards that expose the vast majority of these settings.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 10:37 am
by just brew it!
churin wrote:
Does every "Pass" repeat the exact same test?

Same test algorithms, but IIRC some of the sub-tests use pseudo-random numbers which will change with each pass.

As others have noted, RAM failures are often intermittent in nature. IMO the best way to use Memtest86 is to launch it before you go to bed (or before you leave for work), and let it run multiple passes over several hours when you don't need to use the PC.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 11:01 am
by churin
JPForums:
Thank you for your detailed explanation for another reason why the test needs be run for more than one pass.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 11:40 am
by Flying Fox
I can share my own anecdotal experience: I once had a pair of new RAM sticks and memtest failed on pass #20 out of #23, at around the 13 hour mark. Due to different access patterns, in regular use it may show up sooner. Plus, I was going to overclock so I need the RAM to have a smaller margin of error. Since the sticks were new, so I could return the kit and got a new set and not needed to bother with isolating which stick.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 11:41 am
by GTX
If still interested Memtest86+ 5.00 is available in the "reviewer Kit for Intel Haswell" from http://www.cpuid.com/

From the readme;

The long-awaited Memtest86+ 5 is finally here !

For reviewers, here are some tips :

- Press F1 at startup to enable "Safe Mode". Could be useful if you're reviewing an A-Step HSW or an early engineering-sample of a GT3e-enabled CPU.

- Press F2 to toggle the experimental SMT support in case of any problem. SMT is disabled by default on all Xeon/Opteron and any Engineering Sample CPU. You can press F2 to force SMT, but nothing guaranteed !

- [Hidden Mode] Press F3 to enable an alternative way to detect SMT. Only useful on Gigabyte Motherboard at this time.

Note about SMT : our internal tests showed that SMT only speed up the tests but DO NOT help in finding more errors.




Memtest86+ V5.00 changelog
----------------------------

- Added support for up to 2 TB of RAM on X64 CPUs
- Added experimental SMT support up to 32 cores
- Added complete detection for memory controllers.
- Added Motherboard Manufacturer & Model reporting
- Added CPU temperature monitoring
- Added enhanced Fail Safe Mode (Press F1 at startup)
- Added support for Intel "Sandy Bridge-E" CPUs
- Added support for Intel "Ivy Bridge" CPUs
- Added support for Intel "Haswell" CPUs
- Added support for Intel "Haswell-ULT" CPUs
- Added support for AMD "Kabini" (K16) CPUs
- Added support for AMD "Bulldozer" CPUs
- Added support for AMD "Trinity" CPUs
- Added support for AMD E-/C-/G-/Z- "Bobcat" CPUs
- Added support for Intel Atom "Pineview" CPUs
- Added support for Intel Atom "Cedar Trail" CPUs
- Added SPD detection on most AMD Chipsets
- Enforced Coreboot support
- Optimized run time for faster memory error detection
- Rewriten lots of memory timings detection code
- Corrected bugs, bugs and more bugs

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 12:02 pm
by just brew it!
GTX wrote:
If still interested Memtest86+ 5.00 is available in the "reviewer Kit for Intel Haswell" from http://www.cpuid.com/

Either I am having a senior moment, or the download link for that reviewer kit is very well hidden.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 12:22 pm
by GTX
My bad...... :oops:

I didn't notice it was gone now......I d/l it about a month ago.

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 12:53 pm
by churin
GTX:
I could not find download link for Memtest86+ v.5.0. Could you help?

Re: Memtest86+ behaves strange

Posted: Thu Jul 11, 2013 1:35 pm
by GTX
Again, apologizes for not noticing link was missing.....

Found another link on XtremeSystems Forums for Memtest only, not the reviewer kit: http://www.mediafire.com/download/8mspcdaoaz74vr3/Memtest86%2B+5.00.rar

gl