Paul G. Allen wrote:
[EMAIL PROTECTED] wrote:
On Fri, Aug 29, 2008 at 11:58:18AM -0700, David Brown spake thusly:
Building the kernel is by far the best memory test I've ever found.
The memtest utility I suggested (memtest86+) is an offshoot from the
original memtest86 that you may have had experience with in the past.
The memtest86+ variant was developed and enhanced by the group at
x86-secret and not the original Brady code. This variant is in heavy use
in the labs at Intel and AMD and both companies provide revision changes
back to x86-secret.
We use it to burn-in 8-way Opteron servers with 256GB of RAM. The
current version of memtest86+ is very accurate, usually isolates the
dimm in question on the first pass. Although with Opterons, sometimes it
is the CPU's embedded memory controller.
Compiling the kernel, especially with a high thread count `make -j16`
can be an excellent way to test an overall system prior to deployment
but such a test has the operating system, kernel, local storage, etc as
added variables and does not properly isolate memory. At least with
memtest86+ you are removing several other variables.
Your other options are a benchtop DIMM tester (very expensive, several
grand) or using a hardware only PCI test card from Ultra-X.
Folks at Intel, AMD, JEDEC and CMTL (as well as several of the ram
manufacturers) use memtest86+ and the Ultra-X cards.
If you are really unsure I have a few Ultra-X cards, if you are ever in
the neighborhood.
--
Best Regards,
Jeff Johnson
President
Western Scientific, Inc
[EMAIL PROTECTED]
http://www.wsm.com
5444 Napa Street - San Diego, CA 92110
Tel 800.443.6699 +001.619.220.6580
Fax +001.619.220.6590
"Braccae tuae aperiuntur"
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
