How do I fix this memory problem with my G53?

[bitt3n]

I put 4 4GB sticks in my G53, and memtest86 is giving me errors. CPU-Z says they're running at 1.50V, CL/tRCD/tRP/tRAS of 9/9/9/24. Unfortunately I'm running twinned pairs, Corsair in slots 1/3 and Micron in slots 2/4.

I'm wondering what's the simplest way to fix my problem. Is changing the voltage likely to help (and if so, what should I try), or do I need to resort to pulling out one of the pairs to see if the fact the pairs are mismatched is the issue?

[cl-scott]

memtest86 tends to look for read/write errors, so if you're getting errors then you need to do the time consuming and tedious task of testing the various memory modules as well as the memory banks on the motherboard. If the RAM is bad, replace it, and if it's the motherboard that has bad memory banks you have to weigh the various options open to you.

[bitt3n]

cl-scott wrote:
memtest86 tends to look for read/write errors, so if you're getting errors then you need to do the time consuming and tedious task of testing the various memory modules as well as the memory banks on the motherboard. If the RAM is bad, replace it, and if it's the motherboard that has bad memory banks you have to weigh the various options open to you.

Ok. Is it conceivable that the problem could still be the mismatched pairs, or does the fact memtest86 shows errors conclusively indicate the problem is either a bad stick or a bad mobo slot? (ie, if I remove one pair and see no errors, I'm wondering if I can be sure that one of those sticks is bad, rather than that those sticks are fine, but they can't be mismatched with another pair).

I assume there's no way to disable certain slots to identify which stick is bad without removing the RAM? I can see no obvious way in memtest86 to test individual sticks, which seems kind of unfortunate.

[fostert]

cl-scott wrote:
memtest86 tends to look for read/write errors, so if you're getting errors then you need to do the time consuming and tedious task of testing the various memory modules as well as the memory banks on the motherboard. If the RAM is bad, replace it, and if it's the motherboard that has bad memory banks you have to weigh the various options open to you.

Yes, you must remove the sticks and retest with various configurations (singles, pairs, singles in different banks, pairs with switched banks, etc) until you zero in on the affected stick. Be sure to run each config for 3 full passes of memtest before pronouncing it error-free and moving on.

It could also be that the sticks need to be moved around a bit to make better contact with the slot's pins; sometimes oxidation forms on unused pins and it takes a couple installs to scrape off.

Upping voltage will often solve an error problem, yes, thats how overclockers do it! But does you BIOS allow you voltage control? Remember: higher voltage equals more heat generated by the IMC (on the cpu die) and hence a hotter laptop. However, your sticks should be guaranteed to work at standard DRAM voltages (1.35-1.5V) and if they are throwing errors, you should just RMA them instead of over-volting the modules.

One thing with memtest: it will throw errors even if the sticks are good, but the CPU's memory controller isnt. If you can't find a configuration that doesn't produce errors, it could be the CPU itself. Thats bad and only RMA-able.

[xeromist]

Yes, you would have to remove the memory and test sticks independently to see if any are bad. The way that I would do it would be to remove all of the sticks from the user accessible compartment and run the test to see if you get errors. If not then add one or two sticks at a time and re-run to find bad sticks and/or slots through the process of elimination.

[Zygomorphic]

Mismatching shouldn't generate errors, only cause single-channel mode to kick in or else a drop in frequency/increase in timings or some combination of the two. Errors are from slots or RAM cells, not RAM config.