m-am lovit si eu cindva de problema asta….
freezing cind nu te astepti cu “kernel panic”
am luat-o din aproape in aproape si s-a dovedit a fi o chestie banala (acum nu stiu daca asta o fi hiba si la tine…) la mine a mers dupa aia cu aceeasi memorie deci nu ea era de vina:
curata soclurile in care se infig baretele cu spray pt contacte electrice… 
tot asa , am testat ce Memtest si cu Rember (ruleaza cu boot normal dar nu testeaza decit o anumita parte a ramului)
uneori lucram 2-3 ore si nu se intimpla nimic alteori crapa la 10 minute…
dupa “operatie” nema prablem…
later edit:
poate te ajuta, e de la documentatia lui memtest, ideea cu sprayul mi-a venit dupa ce am citit asta:
TIPS FOR ISOLATING DEFECTIVE DIMMS
Memtest currently does not have the ability to isolate which DIMM or DIMMs are marginal or defective when the test results report a failure. This feature is planned for a future release. For now, the best way to isolate the offending DIMM(s) is to use a binary search methodology. This is an algorithm which is popular in many sorting routines and can lead to the discovery of a defective DIMM in a minimal number of swap/test sequences.
When memtest reports one or more failures, the first step in isolating the offending DIMMs is to remove half of them and then rerun memtest. If there are no failures, then the suspect DIMMs are the ones that were removed. If failures are still reported, then one or more of the still-installed DIMMs are bad.
If the failures are in the still-installed DIMMs, again remove half of them and retest. If the failure are gone, then swap the installed half for the removed half and retest. Each time a test is run, either cut the number of installed DIMMs in half (for a failure) or swap the installed DIMMs for the removed ones (no failure) until the minimal number of DIMMs are installed (e.g., one pair in the G5 systems). Once you’re down to the minimal installation, put back in all of the now known good DIMMs and swap one of the remaining suspects out for the next test. Under normal circumstances, you should be able to isolate the failing DIMMs in just a few swap/test cycles.
There are many other reasons DIMMs can appear to be bad. Sometimes, a defective DIMM socket is the culprit and simply not using that socket solves the problem. Problems can also arise from mixing and matching different brands of DIMMs, especially if they aren’t all rated with the same timing specs. It’s even possible that the motherboard CPU caches may be bad and the fault doesn’t lie with the DIMMs at all. Suspect the CPU caches if the tests pass in single-user mode but fail in a terminal window under the full OS. The CPU caches are turned off in single-user mode and are therefore not part of the memory test, whereas the caches are in the test path under full OS operation.
http://www.memtestosx.org
