increasing voltages to improve ddr3 1600mhz stability

[er557]

I have recently ran hci memtest for windows(8 instances x 1300MB), and it found one error with one thread. Ram is gskill 1600mhz 12gb RL kit 9-9-9-24 1.5v, on asus p6t deluxe latest bios - 2209, i7 965 3.2ghz ,nothing overclocked, triple channel.

This utility is more stressing than memtest x86 for dos, as in windows memory is used differently.(BTW memtest x86+ two passes OK, as have all other 7 instances in hci memtest)

I do have occasional video driver crashes in 2d tasks + corruption on the screen, very rare, nothing else is wrong in any other aspect incl. games and benchmarks. This did start when I upgraded my gtx 280 to a gtx 580, the system was very stable before(though it still may be a manifestation of some other instability, not the video card itself). I troubleshooted that issue as per the video card, and not any tweaks helped, plus I can't RMA it as the issue is not reproducible in lab conditions or in any games or benchmarks).
Occt cuda GPU memtest(video ram test) passed fine.

So I turned to check the pc ram as it may be a cause of a TDR.

So which voltages should I adjust in the bios before testing again(everything is set to auto right now, XMP , 1600mhz, 1.5v.
should I adjust DRAM voltage or QPI/VTT voltage(how do I know the current AUTO value?), and by how much?
Maybe reduce frequency from XMP 1600mhz to 1333mhz, as some say XMP is already overclocking? Will that affect real world performance?
Or can I safely discard the hci error and judge by memtest86+ that my ram is stable?
(The first ever TDR on my system began BEFORE I upgraded the ram from a 6gb XMP kit to an identical 12GB kit- so still with a 6GB kit- and never happened with the 6gb kit and my old gtx 280, for nearly two years.)
Maybe I need to think of upgrading from vista x64 sp2 to 7 x64? BTW I disabled aero and d3doverrider for now, and for now things seem normal(except a gpu crash every month or so and that one hci error).
Thanks!

 

[Goodman]

My Ram is 1.650v but can take 1.8v 24/7 no problem so you should be safe up to 1.650v , you can try little bump at the time start with 1.550v & test

 

[cdawall]

Bump it a little 1.525v and see if it fixes it. It could be something simple like the board not having a stable volt reg for the mem.

 

[er557]

UPDATE:
I just happened to check my drives with hdtune, and one of them showed some pending sector count, and one of my page files is there, which could become sometimes corrupted and send bad data to ram or gpu, causing a tdr. So for now I will remove page file from there, check the drive and see what gives for the next few weeks.

edit: wouldn't have errors would show in two passes of memtest 86+ if anything was unstable RAM wise? They didn't, so I take windows hci test with a grain of salt.

 

[AsRock]

Bump it a little 1.525v and see if it fixes it. It could be something simple like the board not having a stable volt reg for the mem.

Yeah bump it up to a safe voltage and check in the bios what the voltage actually is ( after reboot ) as some it's a littler higher or even a little lower.