F3-12800CL9D-4GBNQ X2 (8GB) + GA-MA790GPT-UD3H + Phenom II X4 965 (RB-C2)

[DeUltima]

Model of the motherboard: Gigabyte GA-MA790GPT-UD3H (BIOS V. F4d)
Model of the memory: F3-12800CL9D-4GBNQ X 2 (4 Sticks / 8 GB)
Model of the CPU: AMD Phenom II X4 965 (Stepping 2 - RB-C2)

I originally built this rig a few months ago with only one F3-12800CL9D-4GBNQ 4GB kit and she ran perfectly with no problems and completely stable without playing with the bios at all.

Being happy with everything except for my system's hunger for RAM, I decided to go out and buy and an extra F3-12800CL9D-4GBNQ 4GB kit. As soon as I plugged in the extra sticks and fired her up, I started getting random BSOD with Memory Addressing/Interrupt Exception errors.

I did have AMD Overdrive installed and had automatic overclocking turned on. Thinking this might be the cause of the problems, I turned everything back to default and uninstalled it. This did help a lot, but I've still been the victim of random BSOD here an there, pointing at the memory as the cause.

After doing my research, I think I'm simply a victim of the flawed memory controller in my prized CPU. And I need to go into the BIOS and manually take control of the situation to stabilize my system and peace of mind once again.

I'm looking for the following settings please:

CAS# latency:
RAS to CAS R/W Delay:
Row Precharge Time:
Minimum RAS Active Time:
1T/2T Command Timing:
TwTr Command Delay
Trfc0 for DIMM1:
Trfc2 for DIMM2:
Trfc1 for DIMM3:
Trfc3 for DIMM4:
Write Recovery Time:
Precharge Time:
Row Cycle Time:
RAS to RAS Delay:

DDR3 Voltage Control:
SB/HT Voltage Control:
NB Voltage Control:

Finally, should I stay at the default 1333Mhz or down-clock myself to 1066Mhz?

[Stilly1155]

This thread might help you.....

http://gskill.us/forum/showthread.php?t=4342

Tim

[DeUltima]

Thanks, but the CPU in that thread is a C3 and has an improved memory controller from what I'm reading. I'm dealing with the older flawed C2, and was looking for stable numbers for the C2 specifically.

[Stilly1155]

No problem....Sorry it didn't help you.

[DeUltima]

That's cool... I'm just wishing I held off on buying my CPU until the C3 version came out. But as they say, When life gives you lemons, make lemonade.

[GSKILL TECH]

DDR3-1333 8-8-8-24 timings will probably be the best with your combination.

So try:

CAS# latency: 8
RAS to CAS R/W Delay: 8
Row Precharge Time: 8
Minimum RAS Active Time: 24
1T/2T Command Timing: 2T
TwTr Command Delay- auto
Trfc0 for DIMM1: 160ns
Trfc2 for DIMM2: 160ns
Trfc1 for DIMM3: 160ns
Trfc3 for DIMM4: 160ns
Write Recovery Time: 10
Precharge Time: auto
Row Cycle Time: auto
RAS to RAS Delay: auto

DDR3 Voltage Control: 1.50V
SB/HT Voltage Control: 1.40V
NB Voltage Control: 1.40V

Thank you
GSKILL SUPPORT

[DeUltima]

UPDATE:

Unfortunately I'm still having troubles, so I wanted to share what I've done since then...

After I used your recommendations, my system went BSOD almost right after boot. I decided to run the Windows 7 Memory Diagnostics Tool. It detected hardware problems right away. Thinking it might be a bad stick, I decided to run each stick one at a time. But all 4 sticks passed no problem on their own. I then put all 4 sticks back in, except now I put the new pair in slots 1 & 2 and the old pair into slots 3 & 4 (Originally I had the old pair in 1 & 2 and the new in 3 & 4). I ran the test again and it passed with all 4 sticks. I setup everything with your numbers again, and everything seemed fine for a few days.

Then while I was trying to reinstall some software, boom I get hit with another memory related BSOD. I ran the memory tool again, and once again it found hardware problems. Even after rebooting a couple times and running the test again. But once I powered down completely and powered back up and ran the test, the memory would again pass the tests with flying colors. So whatever the issue is, soft boots don't seem to be clearing it, it's not until it's power cycled does it clear.

I've had about 3 more memory related BSOD since then, and at a loss to explain what the issue is here. One thing I have noticed however is that CPU-Z does not give the same numbers for my two pairs of RAM, and I was wondering if that could somehow be the issue at all; Possible not syncing / timings slightly off? Take a look at some screen shots I took while in CPU-Z:






I first noticed this when I switched the pairs around, because before the numbers were reversed respectively, and the old pair also are displaying a "SPD Ext." of "XMP" where as the new pair doesn't show a value for "SPD Ext." at all.

Any advice would be greatly appreciated.

Thanks,

[GSKILL TECH]

Send them in for new replacements. Sounds like a module is not initializing properly.

Thank you
GSKILL TECH

[RQuallsJR]

Raise your DDR voltage to 1.65 and use the setup that the GSkill tec gave you

[jallen9]

DeUltima,

Did you ever get this working? I am having a similar issue. I am running:

-Gigabyte MA790GPT-UD3H MB
-GSkill G.SKILL 4GB (2 x 2GB) 240-Pin DDR3 SDRAM DDR3 1600 (F3-12800CL9D-4GBNQ)
-AMD Phenom II X3 720 2.8GHz Socket AM3 95W

Everything has been solid for over a year. I recently added 4 more gig of memory, using the exact type (F3-12800CL9D-4GBNQ) and now I am getting random BSOD's. I ran memtest for 12 hours with no issues.

Anyway, just wondering if you ever figured it out.

[DeUltima]

I thought I did, let me update you on my experience so far:

I tried the suggested settings by everyone with little success, but did notice the sticky post in the DDR3 forums here called "AMD CPU / Memory Controller Limitation".

I searched through the AMD Errata list and found Errata #379 "DDR3-1333 Configurations with Two DIMMs per Channel May Experience Unreliable Operation" and basically the work around was to de-rate DDR3-1333 system memory to 533 MHz operation (DDR3-1066) in your system BIOS. And considering I had the RB-C2 Revision of the Phenom II X4 965, listed as affected by this problem. I gave it a try, and let the BIOS auto timings take over: (CL: 7, tRCD: 7, tRP: 7, tRAS: 20, tRC: 27, CR: 1T) It actually seemed to fix my problems and all my random BSOD vanished, I no longer experienced any noticeable problems until I updated my CPU.

I have now updated my CPU to a Phenom II X6 1090T, I was hoping since the noted errata didn't include the newer CPUs I would be rid of the problem and be able to take her back up to 1333Mhz. Everything went fine for about a week, with no problems. But out of nowhere I got a BSOD, but blew it off because I am overclocking the CPU to 4.0Ghz.

I slowly started getting lots of errors in windows, and Windows 7 suggested it was the memory via the Action Center, so I went and de-graded it again back to 533 MHz operation (DDR3-1066) in the BIOS, and I've been clear of any BSOD since, but I am now getting that behavior from before where on the odd cold boot (Power cycled completely off then back on again) where when Windows boots I get a ton of errors and need to Power Cycle the system again to clear it.

For all I know, that behavior never vanished in the first place, as my computer stays on 24/7 and that problem doesn't happen when rebooting, only when completely turned off and back on again. I can't speak for sleep mode, as I don't use it often, but can't remember any memory problem from just coming out of sleep either and assume since power is still going to the memory while in sleep mode that it wouldn't; However, hibernate maybe another story.

So ya, that's my current work around for all this: Run the memory in 533 MHz operation (DDR3-1066) in the BIOS and never power cycle it unless needed. And if the memory goes wacky after a cold boot, I just power cycle it until the problem vanishes, as it's very noticeable as you boot the OS and start getting errors and services/programs crashing off the bat.

I may take G.Skill up on replacing my sticks, now that I know it's not limited to my old CPU and could very well be that a module is not initializing properly as the Tech suggested. Maybe two at a time, as I don't want to totally lose my system while waiting for my replacements.

I don't know if that helps you at all, but if I ever get this totally solved, I will post my solution here.

[MikeH]

Please refer to my other post re using MS WINDIAG v0.4 from a boot CD and not just relying on MEMTEST86+ alone.

http://oca.microsoft.com/en/windiag.asp

The biggest plus is it allocates error counts to specific stick and I found it showed me errors when MEMTEST86+ was saying all clear. Especially when you hardly ever shutdown the pc. I was finding MEMTEST86 only showed my problem when the pc was cold start but WINDIAG showed it anytime it existed. Huge breakthrough.

I'd also suggest, when you run diag progs to make sure they run immediately upon exiting Windows and do not enter bios in between.

Please reply if you try any of this.

http://www.gskill.us/forum/showthread.php?t=7010

Cheers.

[DeUltima]

Thanks MikeH, but windiag only tests up to a max of 4GB, and I'm dealing with 8GB.

What I'm reading from all my research into the subject is that AMD Phenom II CPUs don't "officially" support four sticks of DDR3-1333 or DDR2-1066. Internal testing showed that four sticks triggered instability in multiple test platforms when using standard JEDEC voltages and speeds. Two high-speed sticks on a single channel can push the memory controller’s limits; combine that with subtle variations in motherboard architectures, and inter-component signaling can become erratic.

Downclocking is your safest bet, but from what I gather if you find the right voltages for the DRAM/North Bridge along with the right timings, it is possible to achieve 1333Mhz . After downclocking, I no longer received any more BSOD and my system runs 100% stable, with the exception of the cold boot problem I explained earlier. I'm not sure if the two are even related, but that too never happens with only 2 sticks installed.

If you truly want to achieve 1333Mhz or higher with an AMD CPU, I would suggest you forget about trying to run 4x2GB and seek out a 2x4GB kit. AMD CPUs have no problem running 2 sticks at these speeds, it's when you add the extra two do you see stability problems. We will just have to pray the upcoming Bulldozer will solve this problem for good.

As these problems seems to be a little more common among some of G.Skill's Dimms like mine and the Ripjaws, I really wish they would take in my 4 sticks of 2GB in trade for 2 sticks of 4GB. I think that's the only real solution in my opinion. The main reason I have not RMAed them, is because I'm not convinced there is anything really wrong with them, that replacing them will solve. I would just have 4 sticks of the same ram, and more likely than not, have the same problem.

[jallen9]

It seems as if I am in the same boat as you DeUltima. I had no idea when I purchased this additional 4gb ram I would be causing myself so many headaches. I think you are right about running 2x4gb sticks instead of the 4x2gb sticks of memory. An even swap from GSkill would be nice.

[GSKILL TECH]

Well another thing is two seperate kits may not be fully compatible with each other. This is why we have tested 8GB kits (4 x 2GB). Sounds like the problem DeUltima is having.

You can request to have the memory tested before shipping to insure they work together. Or ask the RMA department to see if you can exchange them for 2 x 4GB.

Thank you
GSKILL TECH