Re: utility for testing ram?

Doug Ledford (dledford@dialnet.net)
Mon, 28 Sep 1998 23:30:18 -0500

Henrik Olsen wrote:
>
> On Mon, 28 Sep 1998, Ricardo Kleemann wrote:
>
> > Hi,
> >
> > Anyone know of a utility to extensively test a system's ram, in order to
> > determine whether the ram has any faults?
>
> The classic test is to compile the kernel repeatedly, as mentioned several
> times before, this will give the machine a thorough workout, including
> running at close to 100% cpu use for a long time, making for no cooldown
> in idling, which will make marginal components even more likely to fail.
>
> A simple script to do continuous testing would be:
>
> #!/bin/sh
> while true
> do
> make clean
> make
> done
>
> Start it running overnight, if it's still running when you wake up, your
> memory's likely to be ok.

No, no, and NO! If you want to test your RAM, you can't run some test that
is CPU power limited. You'll never access your RAM here faster than the CPU
can compile the kernel, and I got news for people out there. There ain't no
CPU yet that compiles a kernel faster than your RAM can read/write those
source code and object code pages. This is a good CPU test, not a good RAM
test. If your RAM fails during this test with Sig11's or whatever, then it
really wasn't marginal to begin with. I know I've posted this test to the
list before, but without someone posting a better test, I still claim that
your best memory tester that exists is this script:

#!/bin/sh
cd /tmp
tar xzf linux-2.1.123.tar.gz
mv linux linux.save
for i in 1 2 3 4 5 6 7 8 9 10
do
tar xzf linux-2.1.123.tar.gz
diff -U 3 -rN linux.save linux
done

If that script spews anything to the screen, you've failed your memory
test. The only exception to this is if your disk sub-system doesn't use
DMA, then this test is not as good as it could be, but if your system uses
DMA (such as a decent SCSI controller, or DMA IDE) then this test will show
bad RAM much faster and more reliably than compiling a kernel.

-- 


 Doug Ledford  <dledford@dialnet.net>
  Opinions expressed are my own, but
     they should be everybody's.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/