Testing your RAM with Memtest86+

What you're really looking for is "are there errors, or not?" It's really a binary question. Since you found them, you've got errors. Also a good point about the random tests.

 

The previous two posts are right. Errors are errors, and it does not matter how many you have in this case. It just means that specific bits have problems under some conditions, of which MemTest86+ randomly spot-checks as many as it can.

 

Hi.

I normal use Memtest86+ to test my laptop, with no errors, but after reading that Memtest86 is more up to date, I downloaded Memtest86 v3.4 to a floppy disk.

but when i run v3.4, it runs for a few second then half my screen goes black and i get loades of errors, in the moving inversions test, and am forced to reset.

So i download the previous v3.3 and tried that and i get no errors.

So what gives?

regards

John.

 

Hmm, hard to say. I would guess that any errors produced by either software would be not what you want. Memtest86 has been updated it looks like, but it's hard to say which one is better. If you still have the warranty, I would probably just get them replaced. Or maybe post a message to the memtest86+ forums and see what they have to say.

 

Perhaps you should verify that your copy of Memtest86 isn't corrupted. Floppies aren't exactly reliable, so it's possible that it's Memtest itself that was messed up.
Verify that the files on the floppy are *exactly* identical to what they should be (you can use the fc command from the command line to compare files)
If they're different, Memtest was corrupted,and you can't trust its results.

If it's not corrupted, it seems to have found some errors.

Like I, and several others have said already, neither Memtest86 or 86+ are guaranteed to find errors. They perform some random spot checks, and nothing more. So errors that are found in one pass, may not be found in the next 300 passes. But the errors are still there.

 

Just want to clarify here... memtest performs an exhaustive test, which is why it takes so long. However, one of those tests is a "random pattern" test. The random test can sometimes show errors using one pattern, but not another one. One might say that for that particular test, it is using spot checks, but for the overall test regime, I would not say that.

 

Just to clarify your clarification, the "exhaustive" test isn't.
It's still just a spot test.

It tests a few patterns that are likely to reveal errors. It can not test every possible one.

The thing is, if you're lucky, the error is something simple, like "Cell #244532 always returns a 0". If you're slightly less lucky, it might be "Cell #436742 always returns the same as its neighbor". Or maybe "Cell #21547 always returns 0 when next row has not been queried recently, and the previous two cells were read immediately before".
Those are all fairly simple patterns that Memtest can try to test.

But there's no rule that the errors must always be this simple. They can be any arbitrary combination of values in completely different memory cells.
On a 2GB RAM stick, you have 16 billion (2^36) bits to test. Now, if we have to test all combinations of two bits (Cell X only fails when Cell Y contains some value), then you have 16 billion * 16 billion-1 possible combinations. That can be tested too, if you give it enough time.
What if the error only reveals itself if the 10 different cells have a specific configuration?
Then we're up to 16 billion ^10 combinations to test, which is..... a big number. You might be able to finish the test before your computer crumbles to dust, but I'm not sure.
And of course, this doesn't take into account that errors may only occur *sometimes* (3% of the accesses to this cell will report the wrong value)

No application, no matter how cleverly written, can ever perform an exhaustive test. Even the biggest tests they do are only spot checks on the most common patterns.
There are just too many possible patterns, you can *never* test them all.

 

Hi.

I have a new system, details in my sigiture.

I have used Memtest86+ and Memtest86, latest versions from floppy.

Also Vista memory test.

Both Memtest86+ and Memtest86 give me 2 errrors in test 5, and when it reaches 98% the system freezes.

Vista memory test also gives me an error in hardware detected, but not every time.

The chipset type is not shown in Memtest86+ or Memtest86.

I have had my new laptop since last friday and have had no problems with Vista, no lockups, and i have played doom3 and crysis demo without problem.

I have had my laptop running for many hours without problems.

I have removed the sodimms and replaced them in reverse order, making sure they are seated well, but i still get the errors.

Should i be worried.

thanks

John.

 

Well, the errors definitely seem to be there, so yes, strictly speaking you should be worried. However, it may not cause too many problems in your daily use. (Perhaps the memory error only occurs under specific circumstances, or perhaps it's in an area of memory that Windows doesn't often use, or...)

The errors are there (and if it's covered by warranty, get the RAM replaced), but if it runs, it runs. Expect the occasional crash though (or corrupted data, which can be a lot nastier)

So up to you whether or not to be worried. It sounds like the system runs ok under normal usage, but the errors are there, and sometimes they will cause problems.
If you can get it fixed, do so.

 

I recommend you get the memory replaced. There are definitely errors, and eventually there will be problems because of it.

It always sucks to have to deal with this on a new system -- it's always nicer to take it home and live happily ever after, but sometimes it just doesn't happen. Eventually you will start having problems with the system, and they could be caused by this memory or by something unrelated, but you will always wonder if this is the cause. You may be able to get them to ship you new memory before sending back the old.

Also, by getting it replaced it costs the company money, which forces them to try to make a better product in the future. If consumers just accept anything a company gives them, quality goes down the tubes and we all suffer.

 

Hi.

I have just tested both 2GB sodimms separately in both slots, one at a time and i get no errors.

I used Memtest86+ 2.0 to level8 and also no errors with Vista memory tester.


With 4GB installed Memtest86+ 2.0 gives errors at level 5 and stops at 98%

So i only get errors when using 4GB both 2GB sodimms together.

so what is going on???

there is a new bios avialiable v1.20, I am currently using v1.16.

regards

John.

 

I think you are avoiding the inevitable by not returning the RAM, but you raise a good question. You will probably be able to get a more detailed answer if you post to the memtest forums.

 

Well, it can't hurt to try the new bios, but I agree with orev.

As I said on the previous page (or was it the one before?), memcheck86 only makes certain spot checks. It checks patterns that will *often* detect errors, but it's not a completely thorough test (it'd be impossible to finish such a test this side of the end of the universe)

*If* it detects an error, you can be pretty sure there *is* an error.
If it does *not* detect an error, you can't be sure of anything.

If I had to explain your specific problem (only occurs with both dimms installed), perhaps the voltage drops a teeny tiny bit when the motherboard has to run both dimm's, and one of the dimm's then becomes slightly more sensitive to errors. (Yes, I just made that up, so no clue if it's at all plausible. But sounds good to me )

Impossible to say, but if it finds errors, it's probably right. It's only if it finds nothing that you have reason to doubt it.

 

I just want to say that I have an ACER notebook and it is not compatible with memtest!

Tested my old memory with it and it showed errors, tested my new memory and it also showed errors and the laptop is working great!
Goldmemory is showing nothing...

 

If you run into this issue it would be a good idea to test the memory on another system (assuming you have another laptop as you also have old laptop memory). It's certainly possible you have run into a compatibility issue, so it's good to try to double-check.

One thing to watch out for though... the fact that your laptop is working fine is not an indication that there are not problems with the memory, just that the system may not be currently using the areas the might be damaged.

 

I got a similar thing with my 2 x 2Gb modules. I tested them individually in both slots and they both run through fine. When I put them in together they come up with 1 error.

Any Ideas ???

 

Hi.

have you tried the built in memory test in vista, type "memory" in search to run it.

most of the time it finds a hardware error, which it say it will show you details of
but it does not.

My computer is working fine, and it has been running sometimes for 24 hours or more without a problem.

So i am not worried.

Regards

PS. what chipset are you using?

John.

 

My machine is in my Signature..... M1530 Laptop with Intel 965 chipset i reckon.
Only tried memtest v3.3 & 1.7 (should really try the latest versions).

I've not tried the test in Vista, I'll try this when I get home from work and let you know if it throws up errors.

 

Hi.

So your chipset is the same as mine, very strange.

I have tried all the latest memtest+ , I always get errors in test5.

try this is the latest version

it`s the only one to support the 965 chipset.

regards

John.

 

So? That probably just means that the errors are there, but in rarely-used parts of RAM that don't (usually) cause crashes or other noticeable errors.

It certainly doesn't mean that your RAM is fine.

 

Thanks

I was using an older version of memtest.

Tried this and it seems that this was the issue, no errors are reported now with both modules working together.

 

Sigh.

I can't figure out how to stop memtest86+

I used it on my desktop a couple months ago and stopped it then, though I don't remember specifically how.

Right now it's running on my new notebook. I can't find an answer anywhere on google as to how it is to be stopped. ESC simply restarts. And I can't just eject the disc because my eject button won't function, possibily because there's a driver for it that doesn't load when memtest boots?

Someone please tell me the obvious answer as to how I can stop memtest and eject the disc. I'd prefer not to pry my disc drive open.

 

There is no way to stop it other than ESC or rebooting.

I highly doubt that your CD drive is not ejecting because of a driver issue. Hold down the eject button for a few seconds. If that doesn't work, unbend a paper clip and stick it in the small hole right next to the eject button. It may go in an inch or two, and press it somewhat firmly. That should cause the CD to eject.

Otherwise, when your system reboots, go into the BIOS and change the boot order so the disk drive comes before the CD drive.

 

Thanks for the quick response Orev. At the time of my post I had been unable to find the key to open the BIOS (new notebook, still settling in with it). Eventually discovered it was F2. Thanks again Orev!

 

Hi.

has anybody had a problem installing memtest86+ to a floppy in VISTA.

I get an error with rawrite.exe will not run, I have tried v2.00 + v2.01 without success.

I have tried running it as an administrator.

I have had to use XP on my laptop to write the Floppy.

I have been installing memtest86 and memtest86+ since it first came out to floppies, but i cannot get it to work in VISTA.

I have tried internal FDD on my desktop and external FDD on my laptop, no go!!!

anybody know whats wrong.

thanks

John.

 

Hi.

I managed to write the disk.

but i had to download rawwritewin.exe to write the memtestp.bin to the disk.

So does rawrite.exe have a problem with VISTA?

Regards

John.

 

The original "rawrite.exe" only worked on DOS and windows up to 98/ME. For NT-based systems, you need rawritewin. NT based = NT, 2000, XP, Vista.

 

Few years ago was the first and last I tried to run this.
Booted fine from the cd and test started then pc just rebooted by itself.
After second try noticed it stopped the cpu fan, so temperature reached
thermal shutdown limit b4.
Memory can be fully tested from inside the os too, prime95 ....

 

If that was the case then something was wrong with your PC. Memtest should not interfere at all with the cooling/fans on your system. It's a very straightforward thing. Also, you cannot test ALL of the RAM from within the OS, because the OS kernel and other software will always be using part of the RAM. Because that RAM is in use, it cannot be tested.

I do think it's a good idea to run other stress tests, like prime95, sandra, etc... to stress the other parts of the computer as well, but nothing can beat memtest for a pure memory test.

 

System in sig.

1 full pass

time: 1:00:00 lol

that was the exact time

No errors (thank God)

Cool test

 

Seriously, do a few hours of testing. My D430, which FAILS to work right now because of a embedded RAM chip issue, didn't show up with any errors until about 3 hours into running it. After that, errors just kept coming in.

 

Thanks Greg, I'll do that.

Although I bought the RAM at newegg and got one friend of mine shipped it to me, so, no RMA for me if anything goes wrong (I don't expect so, knock on wood).

 

About to run the test on my M52 w/ 4GB of A-Data RAM. Will let you know the outcome in a few (hours, days, maybe weeks)

DD

 

Over on DFI street some of the ram guru's and enthusiast say that a good pass of memtest is leaving it going for atleast overnight, and a good test is 24 hours. This is the only way to know if your ram can handle susptended use and really has no errors.

Plus I have not used the test in so long I forget the individual test names & numbers but its only 2 of the test that really ever fail or put stress on the system so setting up a custom test loop to only use those 2 test is the best way to go. You will find any errors much faster that way.

 

Ran the test for 7+ hours overnight and got 7 passes out of it with no errors so that is good news. The strange thing is that the NBR forum site is the only thing that is crashing my IE v6.0 which caused me to do this memory test.

Anyone else having this issue?

DD

 

Love to update the guide with that. Do you have a link?

 

This is my brain with years of knowledge speaking, to find the actual forum post would be hard since they will be several years old.

Just know that memtest is not a very aggressive test, it takes a long time to find any real errors. If you get errors right away your ram is totally screwed up and needs to be replaced. However even "good" ram can fail after an overnight test and this is a good sign that under heavy stress like during gaming or certain kinds of applications that your ram may fail leading to data corruption or a X color screen of death.

I found that things like prime95 showed ram errors much faster than memtest, but memtest is good because it runs before you even get into windows eliminating some variables and keeping you safe from data corruption.

I do remember the main test was called something like "large byte move" like test 6 or so, thats the hardest test and the one I usually looped so I could cut my test down from overnight to just maybe an hour when making hardware changes to make sure everything is ok.

 

Should I remove the sticker on the ram, because i left it on.

 

I think it will be fine with the sticker. Contrary to popular marketing, you don;t really need to worry about heat on RAM.

 

I have a 32-bit Vista Home Premium OS (HP dv9500t [email protected]) that I upgraded by adding 2@2GB Patriot memory sticks for a total of 4GB (in part for a later upgrade to 64-bit Vista). I ran the Memtest86+ v2.01 "Std" test set on the combined 4GB memory for 34 passes plus, wall time 29:32:07 without any reported errors. When I run the "Windows Memory Diagnostics Tool" (available on the HP at startup) on the combined 4GB memory, "hardware" problems are consistently reported even on pass 1, but, of course, no detail is given, just a note to contact the manufacturer. I have also been running the Prime95 "torture test" for nearly 4 hours now with no reported problems.

Q: Is the "Windows Memory Diagnostics Tool" likely generating the error because I've got more memory in the system than the 32-bit OS can recognize? Should I just rely on the Memtest86+ & Prime95 torture error-free test results instead and consider my memory as likely to be okay?

Thanks for any opinions.

 

It's hard to say, but as far as I know, the Windows test is relatively new. Those other tests take a lot longer and as a result are more comprehensive.

One way to tell would be to run the Windows test on one module at a time. Take one of them out, then run the test. Then replace the one you took out and remove the other one, and run the test again. Then they will both be testing only 2GB, so the 4GB limit should not be a problem.

If you still see errors, hopefully it will only be on one of them, and at that point I'd try to get a replacement.

 

Hi folks,

I am confused (so what's new ) ... anyway, I have an old Dell Inspiron 8100 that it can have Max 512MB of RAM (2x256MB) ... BTW, I am running Windows 2K on this old lappy ... and I have the following in the Slots

Slot #1 (DIMM A or 0): Toshiba THLY25N01B75, PC 133-333-542, 32Mx64 SDRAM, CL2 (this is the original RAM that came with the unit)

Slot #2 (DIMM B or 1): Dell Memory MT8LSD 3264HG - 133C2, PC133S-333-542-Z, CL3. I understand this memory is made by MicroTech for Dell.

Now here is my problem I am not sure if my laptop sees all of the 512MB.

1) When I run CPU-Z ... it only shows slot #1 ... doesn't show slote #2.

Q1: Why CPU-Z only sees one of the RAMs?

2) When I rum RightMark Memory Stablity Test ... it shows Total Physical 511MB, Available Physical 358MB.

Q2: What does that mean 358MB Avail Physical ... does that mean only sees that much of the memory? If so, is that Normal?

The reason I have not run teh Memtest86+ yet ... I started it but the darn thing is like time comsuning ... and I am NOT sure if it will tell lme if it is running teh test on both of my Modules - DIMM A & B!!!

Thanks,

G!

 

Thanks for the response, orev -- this may be of interest to others as well

"Windows Memory Diagnostics Tool" finds no problems with either of the 2GB modules when they are the only modules in the system but reports "hardware" problems when both are installed the system with the 32-bit OS. As noted, Memtest86+ & Prime95 report no problems on exhaustive tests with both modules (4GB total) in the system. I'm assuming that this is just an oddity of the "Windows Memory Diagnostics Tool".

Q: The 32-bit Vista OS should run without problems with 4GB installed, shouldn't it (even though it won't see all of the memory)?

Thanks again for any info.

 

There's no harm to having 4gb in the 32 bit, aside from it not recognizing the whole of it.

Let us know what happens when you try what orev suggested. Honestly if it were me and prime and memtest thought things were fine and my computer was running fine then I wouldn't stress on it.

 

Just wanna say thank you to Orev.

I left memtest running over night (9 passes).
No error found.
Nice!

 

While this thread has some valid information, downloading the bootable iso of memtest is a waist of a cd when you can use this instead:

http://www.ultimatebootcd.com/

It has memtest86+ inside of it with a TON of other things. My bosses old laptop died and with just that one CD I was able to test everything and find that it was a bad hard drive.

Partition tools, hdd tools, cpu tools, memory tools. Its a much better use of a CD and I consider it a must of any tech persons computer repair/diagnositcs kit.

 

Looks like a useful tool. Thanks for the link.

 

Or you can get a Linux live disk

 

OK guys... We know that there are many tools out there for testing systems. This thread is about Memtest86+, so let's not turn it into a contest.

 

some one of the guys that used the programs to test the memorys should put some pictures for the lazy´s :$