Anyone know a fix for Dual Channel instability

[RGone]

Those sticks may do this>"""Both sticks test fine on Prime95 after 12 hours"""but it ain't in the rig you are trying to run now and if they once did it in the rig you fooling with now, then you need to get back to what once was. That is the point you say memory is prime something stable but it is not right now is it? So it is not prime stable anthing in the current setup. But oh well you obviously don't understand me and I deeyam sure don't understand your logic at all.

 

RGone... :confused: as usual.

[Savant]

Ok I see what your saying. This is the first computer I'm running this memory in. I ran them on prime95 in single channel. Dimm 1 and 2. They're unstable in dimm 2 and 4. I ran each stick by them selves in in dimm 2 on memtest86 and also got them both to test at the same time in dimm 2 and 4 in dual channel mode with memtest86. No errors. Also I did some research and the memory is BH-5. There part number is KHX3200AK2. The A means there BH-5. What else can i tell you so you can help? I only mentioned that they where stable in single channel becuase it means it's not the memory its self that is causing the instability.

 

Edit: And I'm not an expert so if there is anything i'm leaving out please let me know. I'm only trying to help you help me.

[pershoot]

if they are 'A', you have CH-5.

[RGone]

or maybe even something else than CH or BH-5. Depending on when it was manufactured. If CH or BH-5 it certainly needs at least 3.0 volts on it likely to handle dual channel as most of it did back when we ran it so hard on the DFI NF2 boards. The two sticks of 512 for 'sure' BH-5 that I have, take 3.1volts to really settle down and run right.

 

RGone... :confused: as usual.

 

if they are 'A', you have CH-5.

[pershoot]

yup.

 

kingston never made BH-5 with 'A' model # in pc3200 ->

 

reference: http://forums.extremeoverclocking.com/show...?t=91533&page=1

[Savant]

So if they are this then what is the recommened voltage. I just don't want to burn them up.

HyperX PC 3200AK2 (2-3-2-6-1T) (KHX 3200AK2) Winbond CH-5.

[pershoot]

you should be able to take them to ~3.2-3.3 if u keep them cool with a fan. they are ch-5 on a smaller fab, so they will get a bit hotter then bh-5 with increased voltage (u should actively cool from 2.9v+ though..). ~3.1+ to get past the voltage limiter (if that is holding you back on the modules).

 

edit: you will void your warranty and possibly burn them up quicker if u give more then what kingston allows (i think it is 2.8). thought i'd just throw that in there. i don't want it to sound like i am condoning breaking warranty, etc.

[Savant]

I just tried 3.0 volts and it didn't help. I also noticed with 4 modules in it's in dual channel mode. here's what it show exactly.

DDR 333 CL: 2.5 ,2T Dual Channel, 128-bit

Tcl: 2.5 Trcd 3 Tras 7 Trp 3 (2T Timing)

 

Edit: What I'm wondering is, is why it's in dual channel mode when the memory being used is Dimm 1 Sumsung pc2700 Dimm 2 hyperx pc3200 Dim 3 Kingston Value ram pc2700 Dimm 4 Hyperx pc3200.

 

And it runs stable. Anythoughts why its doing this?

[m.oreilly]

hey bungholio, you are right re: t1/t2. when i purchased my first cg stepping cpu, i was informed via the dfi nf 250 manual that t2 could now be enabled on said mobo for stablity. then through forums, heard you should disable for latency...then, when going through the bios, saw cpc was auto(assumed 2t was now enabled). i had it the other way 'round. the thing is, i got better latency on some benchies when diabled. anyway, the mod is goofy, get a real psu, and yes you are an odd one...

[Savant]

But it benchs at bandwidth at 4193 mb/s not at the 5000+ that I got the one time I got to bench my dual channel.

[Savant]

I just bought this PSU 2 months ago when my old PSU died on my old PC. I can't afford to get anothr just to test it out. I guess I can see if there is a computer store in town that will let me "barrow" one.

[danny5]

Baaa, you dont need 3v with that memory. I am using 2 year old Hyperx 3500 modules in dual channel mode here @ 2.6v. This same memory easily ran to 217 FSB in my Asus A7N8X, and there is no reason why it should run to the same speed or faster on the NF4 U D.

 

For the life of me, I don't understand why everyone around here says to crank up the memory voltage to extremes for any type of memory. That is BAD advice - period.

 

True, some memory needs 3v to run, but many memory modules DO NOT need that much voltage, and will damage them..

 

I just got my Fortron PSU, so I have not had time to play with the bios settings much. But here are the limited findings that I have discovered. My 3500 module timings are 2-3-3-6. When running in auto mode, the board puts the memory timings at 2-3-3-8. @ 2-3-3-8 the Sandra memory score is around 5600. When I tighten the memory back to 2-3-3-6, Sandra scores drop to 3600.

 

It seems that the A64 memory controller likes slower timings. But this is very preliminary investigation, and needs to be loooked into.

 

So, what I am saying, is to run the memory timings in auto, and see if that helps your problems.