PDA

View Full Version : REMBER Ram test: interpretation of results


TheGimp
Oct 3, 2004, 09:45 AM
I have a Rev A Dual 1.8 G5 w/2 GB RAM (1.5 Apple, 0.5GB "PNY Technologies Optima RAM" from CompUSA). All RAM is 184-Pin PC3200 and passed muster with Techtools PRO 4.0.

Just so you don't think I'm type who enjoys running benchmarks/diagnostics in between beating off, I should mention that a couple of my apps have been behaving erratically and occasionally crashing, and its not entirely certain whether this became more frequent subsequent to adding more ram.

After running Rember for two loops, testing about 1.7GB (the max allocable), I got the following summary (notice how it reports "all tests pasts" at the very end, depite many failures along the way):


Memtest version 4.04M (32-bit)
Copyright (C) 2004 Charles Cazabon
Copyright (C) 2004 Tony Scaminaci (Macintosh port)
Licensed under the GNU General Public License version 2 only

MacOS X (Darwin) running in multi-user mode
POSIX version 198808
Pagesize is 4096
Pagesizemask is 0xfffffffffffff000
Requested memory: 1778MB (1864368128 bytes)
Available memory: 1778MB (1864691712 bytes)
Allocated memory: 1778MB (1864368128 bytes)
Attempting to lock allocated physical memory....memory locked successfully

Running 2 test sequences...

Test sequence 1 of 2:
Stuck Address : FAILURE: possible bad address line at offset 0x169e4335.
Skipping to next test...
Random Value : \|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|/-\|FAILURE: 0x6ebd7734 != 0x6abd7734 at offset 0x08cb372d.
FAILURE: 0x304941af != 0x344941af at offset 0x08cb372d.
Compare XOR : FAILURE: 0xe451df17 != 0xe851df17 at offset 0x08cb372d.
Compare SUB : FAILURE: 0x8fd59c9f != 0x73d59c9f at offset 0x08cb372d.
Compare MUL : Compare DIV : ok
Compare OR : ok
Compare AND : ok
Sequential Increment: ok
Solid Bits : setting 0FAILURE: 0xffffffff != 0xfeffffff at offset 0x08cb46fd.
Block Sequential : setting 0FAILURE: 0x1d1d1d1d != 0x0d1d1d1d at offset 0x08ba4335.
Checkerboard : setting 0FAILURE: 0x55555555 != 0x54555555 at offset 0x08abd725.
Bit Spread : setting 0FAILURE: 0xfffffd7f != 0xeffffd7f at offset 0x08ba4335.
Bit Flip : setting 0FAILURE: 0xfffffffe != 0xeffffffe at offset 0x08ba4335.
Walking Ones : setting 0FAILURE: 0xfffffff7 != 0xeffffff7 at offset 0x08ba4335.
Walking Zeroes : setting 0
Test sequence 2 of 2:
Stuck Address : setting 0FAILURE: possible bad address line at offset 0x168fd725.
Skipping to next test...
Random Value : \|/-\|/-FAILURE: 0xfd26e2cf != 0xfc26e2cf at offset 0x08abd725.
FAILURE: 0x02ed7fa8 != 0x03ed7fa8 at offset 0x08abd725.
Compare XOR : FAILURE: 0x38df1f1b != 0x39df1f1b at offset 0x08abd725.
Compare SUB : FAILURE: 0xfb328c4e != 0xc5328c4e at offset 0x08abd725.
Compare MUL : FAILURE: 0x00000002 != 0x00000001 at offset 0x08abd725.
Compare DIV : Compare OR : ok
Compare AND : ok
Sequential Increment: ok
Solid Bits : setting 0FAILURE: 0xffffffff != 0xefffffff at offset 0x08ba4335.
Block Sequential : setting 0FAILURE: 0x57575757 != 0x47575757 at offset 0x08ba4335.
Checkerboard : setting 0FAILURE: 0x55555555 != 0x54555555 at offset 0x089ae6fd.
Bit Spread : setting 0FAILURE: 0xfffffffa != 0xfefffffa at offset 0x08abd725.
Bit Flip : setting 0FAILURE: 0xfffffffe != 0xeffffffe at offset 0x08ba4335.
Walking Ones : setting 0FAILURE: 0xfffffffb != 0xeffffffb at offset 0x08ba4335.
Walking Zeroes : setting 0
All tests passed.


What I'm wondering is:
1) Is Rember intended fortesting G5 64-bit ram? The log mentions "32-bit testing".
2) Has anyone else had similar feedback on their systems, yet no application instability attributable to hardware?
3) Are there other recent utilities similar to Rember, with whose results I may obtain a useful comparison?
4) Lastly, is anyone savy enough to discern in the above log whether the reported failures/errors are detected in a particular module? Remember that it is the RAM I purchased at CompUSA which is the most likely suspect (Apple RAM has a pretty good rep, no?).


Thanks in advance for any informed advice. I don't really *need* the usual slew of comments, to the tune of "failures are always bad (regardless of which util. is reporting)" or "suckah!!! That's what you get for buying cheap ram" (I already told myself that).

varmit
Oct 3, 2004, 11:40 AM
Long story short, its most likely bad RAM, or not compatible RAM if you just bought it at Compusa and it didn't say G5 Compatible on it.

TheGimp
Oct 3, 2004, 02:49 PM
It actually said G5 compatible on it. I plan to remove the ram and run the tests again to see what happens. This kinda sux. I hope some retired guru (the type who leaves long messages quoting different parts of original messages and addresses them point by point) replies. Until then...I guess I'll just run Cinebench with the AC on and see if I can score a few more points.