Neo2 owners...some memory tweaking tricks I may have found

[Duvie]

Here are some things I haved witness from continued memory tweaking and testing....this was brought on by getting a flaky beta bios (modded version) and subsequent several flashes later to older bioses and then back up to 1.8 to get once before stability...It is achieved now...

1) I have found that manually setting timings in the bios IS not the same as letting the bios set them. Case in point...

I set the Bios to 200 divider, 1t then cas timngs 2.5,4,8,4....2.75v

Upon running memtest I got 1 error in the 8th and 12th pass before I stopped it....

I then set the Bios to 200 divider and 1t at 272HTT and then set all cas timings to idle. I was using 2.75v (set).....It as I previous noticed set it to same 2.5,4,4,8...However in A64tweaker I noticed some of the numbers of other timings (that we cannot set in bios) were slightly different. This time the test ran for 36 passes before I turned it off...NO ERRORS

So it seems that we are not apples and apples even though we manually set the same 4 numbers...

I am actually becoming more fond of the auto setting. On my Neo2 board with the bioses I am using it is not SPD timings but rather aggressive timings in my opinion. My spd timings for roughly same DDR speed as seen bu CPU_Z was cas 3-6-6-11...waaayyyy off of what it really did...

I may try setting vdimm at different levels and see if the auto selects something different.

**update** Vdimm set has no effect...



2) I just found this one, however I have tested in a fw times now and got similar results...

I am now at a higher HTT level (279) and testing a 166 divider setting... I let it set the function with auto setting again....This time it came up with same 2.5-4-4-8 with 1t...

Now in memtest setting it errors out 8 to 9 times in first pass....

I use A64 tweaker and I set the tras up to 10...it errors out 5 to 6 times in first pass...helps but very little...

Then I noticed that memclk 2 and 3 (known as slots 3-4) are enabled. I disabled them for sh^ts and giggles and then tried again back at 2.5-4-4-8...It now passed the entire first pass no errors. It did get one eror in the 2nd pass and 1 more in 3rd pass...needs some more fine tweaking but that is a sizeable improvement....

What do you think coul be the cause of this. By disabling them am I somehow turning power off to them and possibly strengthening the signal or power to the remaining 2 sockets???

Others please try this...

I am redoing at 2.8v of vdimm to see if that is what is needed to help the auto settings...

UPDATE #!

2.8v a little worse...i tell you these TCCD chips do not like heat, especially Gskills who are too stupid to put heatspreaders on these things....


I set all setting to Auto except the 166 including 1t and vdimm...It still set it to 2.5,4,4,8, 1t but I dont knwo what the vdimm is.....I have the memclk 2-3 disabled...says dual channel in cpu_z and 1mb in SNM so I know I didn't unactivate one of my memory modules and now running in single channel mode...

GOOD NEWS... on my 3rd pass no errors




***sidenote***

You can save the settings in A64tweaker afterwards and have it apply at bootup....

 

[RichUK]

hmm i might have to try this when i get back home .. WAIT a minute home time is in 1 min WOO HOO .. 5.00pm finish .. report back in a bit

 

[Duvie]

Any idea where I can see what my vdimm is set to other then that virus known as corecenter??? I dont even installl that thing anymore...

sandra says 2.82v so that is odd the 2.8v set in the bios didn't take but auto set it right...confusing if one can trust sandra.....

 

[Vegitto]

And you know, one can't trust Sandra.

 

[Stokes]

Duvie -

Are you basically saying mother knows best in this case? And that the timings the motherboard provides are probably the best ones to go with?

 

[jpeyton]

2.8v a little worse...i tell you these TCCD chips do not like heat, especially Gskills who are too stupid to put heatspreaders on these things....

Most TCCD does not like voltage; they perform better at 2.7V, even at high clocks. I've heard that the newest revision of TCCD, and also the new TCC5, do like voltage, up to 3.0V.

GSkill doesn't put heatspreaders on their memory because heatspreaders typically trap heat and make your RAM run hotter. Their only purpose is to balance the heat load between all the RAM ICs on the stick, so some aren't running hotter than others. That is the official word from OCZ reps on another forum, and I've read many accounts of people reporting their memory doing better with the heatspreaders removed.

 

[Duvie]

I think it would have also been better if MSI and other boards had the dual channel stick in slots 1-3 or 2-4 but not directly next to one another to allow better airflow....

Jpeyton that sound good and all but properly place heatspreaders could help draw heat away from the chips and out onto the surface of the spreader which is usually at least 2x the surface area of the chips themselves...Have good case airflow and this can be more efectively swept from the surface....

I think OCZ is right for the half-arse job I see many of them doing with how the heatspreaders make contact....

in theory it should work the way I have stated, but I dont doubt in some cases what you have seen or been told is also right on the money...

 

[Duvie]

Originally posted by: Stokes
Duvie -

Are you basically saying mother knows best in this case? And that the timings the motherboard provides are probably the best ones to go with?

I guess I am somewhat...this may be very bios limited meaning as they have updated the bios to make some tweaks for certain memory compatability....This may not work as well with say an older 1.6mod bios...I would have to check to see if in all different bioses it autos out to the same timings....

IN this case so far it seems to have done 13 passes now no errors.....that is sizeably better then before so something must do better...


In my case 1 at the 200 divider it was undeniably better. For case 2 setting its own vdimm I will reserve full judgement later but 2nd pass versys no errors in the 13th pass...seems better there as well....


The turning off the memclk case may have to be studied by soem one who can test the sockets to see if it shuts of power to those, increase to the others, or just give more clean distribution of power to the others...i dont have those instruments to do the test that really need to be done....

 

[rise]

Originally posted by: Duvie
Any idea where I can see what my vdimm is set to other then that virus known as corecenter??? I dont even installl that thing anymore...

sandra says 2.82v so that is odd the 2.8v set in the bios didn't take but auto set it right...confusing if one can trust sandra.....

you can see what you're getting here i think. i think thats your board. i was bored so i thought i'd look around a bit.

i check my dfi-sli dr this way and mine at least runs a bit high, like +.04. may be worth looking at.

edit--should you want to try, and i never have, you can also read the individual slots by using the dmm in a particular pin of them. i'd imagine it would be the same on all. i know its pin 7 on mine but i'm not sure of the orientation and how it compares to yours.

i've never been inclined to try.