G Skill RAM went bad, RMA'd .. but still questions.

[WeeHawky SinkAllTheMayo]

So I got some RAM that , when tested together, showed up erroring in Memtest86+ and Windows memory testing.

But when taken apart, one stick tested at a time, it didn't fail. Any ideas on what really happened ? GSkill doesn't do testing at the RMA facility , just replaces, so I had no help from them.

I bought the same kit , so I had no down time, and put it in.. Tested.. no errors. Got the kit back (new from RMA department), put it in with the other new kit I bought and all 4 sticks test no errors.

So I'm still wondering what happened on the other 2 sticks that I bought 3 years ago ? How would they not be bad seperate but together ?

[[AU] Tabris:DarkPeace]

Microscopic electrical differences between the pair in a batch combined with a 1 in 1 million chance of the motherboard / memory controller not liking that pair.

If they are used in the same channel of a dual channel configuration with four sticks in total they might even test fine.

RAM is funny like that.
You were most wise to use MemTest86+ to test it.


It could also be dust on a particular memory slot causing it to give errors.

[WeeHawky SinkAllTheMayo]

Originally posted by TabrisDarkPeace:

Microscopic electrical differences between the pair in a batch combined with a 1 in 1 million chance of the motherboard / memory controller not liking that pair.

If they are used in the same channel of a dual channel configuration with four sticks in total they might even test fine.

RAM is funny like that.
You were most wise to use MemTest86+ to test it.


It could also be dust on a particular memory slot causing it to give errors.

Something as simple as dust could have caused the whole issue ?

[[AU] Tabris:DarkPeace]

Well, dust isn't 'simple' from an electricity viewpoint and RAM is just really tiny capacitors and transistors, so yeah, by RAM standards dust is 'huge'.

It's like having gold contacts on the motherboard, and tin contacts on the RAM... the difference in voltage, ampères, resistance, etc. caused by having two different materials in contact with each other is enough to cause problems.

[WeeHawky SinkAllTheMayo]

Originally posted by TabrisDarkPeace ᴳᵀ:

Well, dust isn't 'simple' from an electricity viewpoint and RAM is just really tiny capacitors and transistors, so yeah, by RAM standards dust is 'huge'.

It's like having gold contacts on the motherboard, and tin contacts on the RAM... the difference in voltage, ampères, resistance, etc. caused by having two different materials in contact with each other is enough to cause problems.

Still though, why would memtest and windows memory test pass the sticks seperate then ?

[_I_]

did you make sure your sing the corret xmp/spd profiles for the ram kit

any ram will give errors if its overclocked or undervolted too far

[WeeHawky SinkAllTheMayo]

Originally posted by _I_:

did you make sure your sing the corret xmp/spd profiles for the ram kit

any ram will give errors if its overclocked or undervolted too far

I still got no idea about these profiles.. Unless you are talking about going in and man setting the settings in BIOS? I did that. It was set to Auto and was using the wrong timing, 9-9-9-24 instead of 8-8-8-24 by default (The 9-9-9 is the wrong set from the sticker).

It was using 1.5 V auto already though, but I went ahead and man set that.

[_I_]

set it to 1333 cl11
then run cpuz and see what profiles the ram has
cpuz -> spd tab
select each dimm to see its profiles

when mixing ram
use highest common speed, with highest timings and voltage
if its an intel 1150/1155 build, do not use profiles with voltage over 1.55v

[WeeHawky SinkAllTheMayo]

Originally posted by _I_:

set it to 1333 cl11
then run cpuz and see what profiles the ram has
cpuz -> spd tab
select each dimm to see its profiles

when mixing ram
use highest common speed, with highest timings and voltage
if its an intel 1150/1155 build, do not use profiles with voltage over 1.55v

I set it to 1333 MHz in the BIOS. Just a note: I updated BIOS to the latest version when trying to see if that was the issue, it wasn't.


XMP 1.1 it says in SPD Ext. Max Bandwidth 667 MHz (Half 1333 ?) .

edit: Also on that tab I notice something funny..

edit2: http://picpaste.com/684b055e9027a2e4d2f2d0036a94ea12.png . The timing still isn't exactly what I set it to in BIOS. Perhaps I need to go back and check that the new sticks didn't reset the settings ?

[_I_]

correct, mhz is half of the speed
666mhz = 1333

that dimm only supports upto 1333 cl8, 1.5v
how about the other ones?

[[AU] Tabris:DarkPeace]

Single channel can have tighter timings than dual channel, and if they just kept the same timings it would not have stressed the memory enough.

[WeeHawky SinkAllTheMayo]

Originally posted by _I_:

correct, mhz is half of the speed
666mhz = 1333

that dimm only supports upto 1333 cl8, 1.5v
how about the other ones?

They are all the same models.

edit: F3-10666CL8D-4GBRM

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231275

[_I_]

if you set your ram profile to 1333 cl8 -8-8-24 and ram voltage to 1.5v you should be set

[WeeHawky SinkAllTheMayo]

Originally posted by _I_:

if you set your ram profile to 1333 cl8 -8-8-24 and ram voltage to 1.5v you should be set

In BIOS ?

[[AU] Tabris:DarkPeace]

Usually in the BIOS yes,
You may need to press [Ctrl]+[F6] to unlock an advanced menu to set the chipset/RAM timings.