Alternatives to MemTest

  • Thread starter Thread starter geoff
  • Start date Start date
G

geoff

I downloaded the latest memtest, created a boot floppy, and restarted my
machine.

Memtest said some memory is over 4 gigs and will not be tested, and to press
'C' to continuer, any other key to exit.

At that point, it was frozen, no response.

Anything else to use/try?

--g
 
geoff said:
I downloaded the latest memtest, created a boot floppy, and restarted
my machine.

Memtest said some memory is over 4 gigs and will not be tested, and
to press 'C' to continuer, any other key to exit.

At that point, it was frozen, no response.

Anything else to use/try?

--g
Click to expand...

Are you using the most current version of Memtest86+?:

http://www.memtest.org/#downiso . This is the download page and lists
the different options for running Memtest86+, current version V2.01
 
I downloaded the latest memtest, created a boot floppy, and restarted
my machine.

Memtest said some memory is over 4 gigs and will not be tested, and to
press 'C' to continuer, any other key to exit.

At that point, it was frozen, no response.

Anything else to use/try?

--g
Click to expand...

If you have Vista it has a memory tester built in. Press F8 during bootup
and I think you will see it there. Or you could boot to the Vista DVD to
run it.

Get memtest86+, I ran it with 4GB and had no issues.
 
geoff said:
I downloaded the latest memtest, created a boot floppy, and restarted my
machine.

Memtest said some memory is over 4 gigs and will not be tested, and to press
'C' to continuer, any other key to exit.

At that point, it was frozen, no response.

Anything else to use/try?

--g
Click to expand...

There are two sites for memtest.

http://www.memtest.org/ memtest86+, based on an earlier Chris Brady version

http://www.memtest86.com/ Chris Brady's own commercial/free effort.

In terms of "reserved" memory areas, memtest may rely on the BIOS memory
map. Some areas in low memory, for example, may be off limits
to testing. On a board with integrated graphics, the frame buffer area may
not be available for testing (for obvious reasons).

In addition to the two above, Microsoft offers a test program for download.

http://oca.microsoft.com/en/windiag.asp

So there are a few to choose from. I've always used the memtest86+ one myself.

Paul
 
I downloaded the current version and the reason the program does not respond
is because I am using a USB keyboard.

I ran memtest86+ for 8 hours, testing 4 gigs of RAM, and it came up with 1
error.

Some people take a hardline and feel an error is an error, replace the
stick.

However, this is a Gigabyte MB, GA-MA78GM-S2H, with an AMD quad core phenom.

I am not sure how willing I am to trust newer hardware with older programs.
Everest Home could not read the CPU temps at all, so, I downloaded Everest
Ultimate (30 day trial) and the cores were at 30c but the CPU temp was over
80c.

I found a beta of Everest Ultimate, newer version, which corrected that
issue. The CPU temp now reads 31c.

In the BIOS, there is a ganged and unganged mode for memory. I think
unganged, the default, gives each CPU access to its own chunck of memory.

The bios also has 'load safe settings' and 'load optimizied settings'.
Would any of this affect the memory testing in a DOS type environment? I
don't know but I am also not 100% convinced yet either.

--g
 
geoff said:
I downloaded the current version and the reason the program does not
respond is because I am using a USB keyboard.

I ran memtest86+ for 8 hours, testing 4 gigs of RAM, and it came up
with 1 error.

Some people take a hardline and feel an error is an error, replace the
stick.

However, this is a Gigabyte MB, GA-MA78GM-S2H, with an AMD quad core
phenom.
I am not sure how willing I am to trust newer hardware with older
programs. Everest Home could not read the CPU temps at all, so, I
downloaded Everest Ultimate (30 day trial) and the cores were at 30c
but the CPU temp was over 80c.

I found a beta of Everest Ultimate, newer version, which corrected
that issue. The CPU temp now reads 31c.

In the BIOS, there is a ganged and unganged mode for memory. I think
unganged, the default, gives each CPU access to its own chunck of
memory.
The bios also has 'load safe settings' and 'load optimizied settings'.
Would any of this affect the memory testing in a DOS type
environment? I don't know but I am also not 100% convinced yet
either.
--g
Click to expand...

Memtest should run without a single error, BUT: it could be as simple as
the memory needing a bit more voltage, or for you to set the memory
timings manually, rather than default by SPD. A true error would show
up every time the "bad" address was tested.

The default voltage for DDR2 memory is 1.8V, yet many memory modules are
speced at higher voltages, which must be set manually. Double check the
voltage requirements of the memory you have and make sure it's set at
the proper voltage in the BIOS.
 
If you are using SPD settings then you have a problem.
If you repeat the test and get the same error you really have a problem.
Click to expand...

One error like that means, for me, if the system were BSOD'g or freezing,
weird things like that then that is a problem.

I've already ran flight simulator, highest settings, and played chess, full
graphics, then back to work, using ms word, and other things with no issues.

--g
 
geoff said:
I downloaded the latest memtest, created a boot floppy, and restarted my
machine.

Memtest said some memory is over 4 gigs and will not be tested, and to press
'C' to continuer, any other key to exit.

At that point, it was frozen, no response.

Anything else to use/try?
Click to expand...

Vista's built-in memory test? I have no idea how good it is.

I've never used MemTest because it received a bad review from
RealWorldTech.com. OTOH I have tried a few versions of MemTest86,
MemTest+, and Gold Memory. Ver. 3.4 of MemTest86 didn't work on a
GeForce or nForce mobo (lower or top half of screen was blank), but
ver. 3.3 ran fine and was one of the best tests at detecting errors.
The other was Gold Memory 5.07, which has usually detected errors a
lot faster than MemTest86 has, and it did better in
RealWorldTech.com's review. MemTest+ ver. 1.70 and I think 1.80 and
Gold Memory 6.xx passed many modules that MemTest86+ v. 3.3 and Gold
Memory 5.07 said were bad.

Does anybody know of a program that uses the same test method that RST
does?
 
geoff said:
I ran memtest86+ for 8 hours, testing 4 gigs of RAM, and it came up with 1
error.

Some people take a hardline and feel an error is an error, replace the
stick.
Click to expand...

Like those idiots who insist that all the links in the chains be
good.

http://en.wikipedia.org/wiki/Silver_Bridge

Yes, a single bad memory bit can be just as deadly.
However, this is a Gigabyte MB, GA-MA78GM-S2H, with an AMD quad core phenom.
Click to expand...

If an error appeared after another 24 hours of testing, I'd replace
the RAM, unless you were overclocking it. After all, RAM has got to
be really bad to fail a test run on mobo and not a on a gazillion-
dollar testing machine, especially when the RAM is run at just 50
Celcius, or way cooler than its rated 85C or 95C.
 
I loaded the safe defaults in the BIOS and it passed all the memory tests.
To be honest, I am not sure what the difference is between safe and
optimized defaults. As far as I can tell, they appear the same (speed,
timings, etc.).

One pass for all the tests took the same amount of time with the safe
defaults, 52 minutes for 4 gigs.
If the error is repeatable then you definitely have bad ram.

If this is the case then as soon as some program data is written into
that area of ram, then it may get corrupted under certain conditions.
Click to expand...

I guess it is a confidence issue. If errors were popping up throughout the
tests then I would think there is an issue.

However, for the more paranoid, what you write cuts both ways, unless one is
willing to re-certify their hardware every week or so, data from your
finance app could be written out incorrectly because a byte went bad two
days after it was tested, or data saved incorrectly because of a bad sector
on the HDD, etc.

I doubt even the NSA re-certifies all their hardware each week. A bank I
did some contracting for had a call center of about 10,000 people, all on
one floor, and they did not certify the hardware each week/month/etc.

Given that, it would be quite possible for them to look at your checking
account balance and find an extra zero there (or one zero less) because a
stick of memory went bad (one byte on it) or a bad sector on the HDD.

--g
 
In message <[email protected]> "geoff"
I downloaded the current version and the reason the program does not respond
is because I am using a USB keyboard.
Click to expand...

Enable "USB Legacy devices" in your BIOS settings.
I ran memtest86+ for 8 hours, testing 4 gigs of RAM, and it came up with 1
error.

Some people take a hardline and feel an error is an error, replace the
stick.

However, this is a Gigabyte MB, GA-MA78GM-S2H, with an AMD quad core phenom.

I am not sure how willing I am to trust newer hardware with older programs.
Everest Home could not read the CPU temps at all, so, I downloaded Everest
Ultimate (30 day trial) and the cores were at 30c but the CPU temp was over
80c.
Click to expand...

memtest is different, it's not a matter of newer hardware vs older
programs, if memtest is reporting an error, that means memtest wrote
data to memory that it could not successfully read back.

That's a problem on ANY hardware.

That being said, there are other suggestions in this thread to help
address it, an error doesn't mean bad or failed hardware, just that
there is a problem.
 
geoff said:
I loaded the safe defaults in the BIOS and it passed all the memory tests.
To be honest, I am not sure what the difference is between safe and
optimized defaults. As far as I can tell, they appear the same (speed,
timings, etc.).

One pass for all the tests took the same amount of time with the safe
defaults, 52 minutes for 4 gigs.


I guess it is a confidence issue. If errors were popping up throughout the
tests then I would think there is an issue.

However, for the more paranoid, what you write cuts both ways, unless one is
willing to re-certify their hardware every week or so, data from your
finance app could be written out incorrectly because a byte went bad two
days after it was tested, or data saved incorrectly because of a bad sector
on the HDD, etc.

I doubt even the NSA re-certifies all their hardware each week. A bank I
did some contracting for had a call center of about 10,000 people, all on
one floor, and they did not certify the hardware each week/month/etc.

Given that, it would be quite possible for them to look at your checking
account balance and find an extra zero there (or one zero less) because a
stick of memory went bad (one byte on it) or a bad sector on the HDD.

--g
Click to expand...

That is why "real" computers use ECC :-)

The mystery, is why it isn't more common on desktop platforms.

Paul
 
That is why "real" computers use ECC :-)

Good advice and something to keep in mind the next time someone here asks
for hardware advice. If they intend to use it for things other than gaming
then ECC is a good choice. I do not think my MB supports ECC. There are no
settings in the BIOS for it and the manual mention nothing about it.

Places I've worked at, based on how bad some of the computers looked, I
would be willing to bet at least 10 to 20 percent could not pass a
hardware/HDD/memory test.

Even in small companies, no one verified the hardware but in large
companies, one building would have thousands of computers, at least.

--g
 
Back
Top