RAM is doing weird things

Hi,

At the moment I am using a Asus P8Z68-V Pro Gen3 w i7-2600K@4,6GHz w 2x4GB Kingston 1333 9-9-9-24 value RAM and a 850W PSU.
The RAM is from my old 1156 system and worked fine at 1600 9-9-9-24 for a long time. Now I put it into my new system and started playing around a little bit with latencies, frequency and voltage. Suddenly I ran into stability issues, did Memtest+86 and found errors in the first RAM slot. I removed the RAM and tested again - no errors. Put the removed RAM back and again no errors. So I kept using it and continued playing with the settings until I again run into stability problems. Strange thing was that even after setting everything to default in the bios the stability issues remained and windows based Memtest immediately found errors. Did Memtest+86, two rounds, no errors at all. Swapped RAMs, and no errors in windows-based Memtest. Now everything runs fine but have no clue what was going on.

The way I ran into stability issues was that I tested both directions; stock frequency, stock voltage with as low as possible latency. Brought it down to 7-7-7-17 T2 and stable with all tests. Aida, IntelBurn, Memtest. Tried to tweak further with T1 --> crash.

Other direction; frequency as high as possible with adjusted latencies and voltage but not higher than 1.6V. Reached 1600 as expected - everything fine. 1866 --> crash.

In both cases these crashes led to the strange behaviour that no matter what setting I used afterwards the system remained unstable until I started taking out the RAMs and playing around with them.

Any ideas what could be going on?

Chris

I know this probably isn't the answer you are looking for but you should probably stick to stock voltage, timings and frequency to maintain stability. The obvious reason is stability but the reality is that performance gains are negligible and probably not even noticeable in real world applications.

Also the JEDEC Standard for that RAM is 1.5V... If you put a lot more volts to it (you mentioned 1.6V which should still have been safe but the upper range is listed at 1.575V) you may have damaged it trying to overclock it.

King107s said:

I know this probably isn't the answer you are looking for but you should probably stick to stock voltage, timings and frequency to maintain stability. The obvious reason is stability but the reality is that performance gains are negligible and probably not even noticeable in real world applications.

Also the JEDEC Standard for that RAM is 1.5V... If you put a lot more volts to it (you mentioned 1.6V which should still have been safe...

Click to expand...

+1 as quoted.

Hi,
thanks for your replies. In the meantime I did a lot of testing. I had the effect that when taking out 2 DIMMs everything was running fine. I tried all DIMM-variations and slot-variations. RAM settings were at factory. It always worked with 2 DIMMs only. As a next step I clocked the CPU back to factory and tested with all 4 DIMMs and it worked fine. So at the moment I have to consider that if I overclock my CPU my system can handle only two DIMMs.
Is this a CPU/IMC- or a motherboard-issue?

4 DIMMs do add stress to the IMC, sure. Did you raise that voltage a bit to try to stabilize it?

As ED said, 4 sticks will add more stress and this is amplified (stability wise) when your timings are on the edge of what the RAM can do.

I would try resetting the CMOS and let the BIOS go back to fail safe defaults and start overclocking with only the CPU, testing the CPU & RAM stability as you go.

Also, as a default, go ahead and override the auto set points in the BIOS for the RAM and manually set it to 1333MHz 9-9-9-24 @ 1.5V.

EarthDog said:

4 DIMMs do add stress to the IMC, sure. Did you raise that voltage a bit to try to stabilize it?

Click to expand...

No havent tried that yet since the whole testing was done with all BIOS settings to auto (which resulted in 1.5V, 9-9-9-24, T2 for the RAM) and manually set factory settings.
Everything worked perfectly for months on my old and much weaker 1156 board, even at 4GHz and 1600MHz RAM frequency at 1.5V and 9-9-9-24.
But I will try it.

@king
I tried both ways. Let mobo set the factory settings and manual setting of the factory settings. No difference. Still instability with 4 DIMMs. What could be the difference between CMOS reset and going back to default settings?

Again, is this a IMC-weakness issue or could a mobo also cause such problems? From my understanding I would suspect the IMC...

I decided to do the CMOS reset (with the CLRTC pins, not the battery method) and for the moment it seems to work - at least with factory settings. (no OC for the moment) I recalled that problems started when I had the iGPU activated in BIOS in addition to my dGPU. Could be coincidence but I decided to deactivate the iGPU to ease overall stress for the CPU.
I will do further testing with OC settings. Lets hope the best.

Thanks for your help!

Hi,
I would like to bring it up again since I kept tweaking my system and observed one constant. What I see is that I run IBT (at the moment at 5GHz) and I pass the tests. I restart to make settings in the BIOS and at the very beginning of startup I get a Memory failure with the MemOK LED blinking. I happens with 2 DIMMs as well as with 4. I am running the RAM at stock (1333, 1.5V). At the moment I tried to increase the RAM voltage to 1.55V and IBT is running at the moment. It seems that everything is ok as usual.

Any ideas what this could be? I am thinking about buying 2x8GB of 1600 RAM to replace my old set but the problem is that if it has nothing to do with the DIMMs, I wasted money because returning RAMs is not possible.

cubusmedusa said:

Hi,
I would like to bring it up again since I kept tweaking my system and observed one constant. What I see is that I run IBT (at the moment at 5GHz) and I pass the tests. I restart to make settings in the BIOS and at the very beginning of startup I get a Memory failure with the MemOK LED blinking. I happens with 2 DIMMs as well as with 4. I am running the RAM at stock (1333, 1.5V). At the moment I tried to increase the RAM voltage to 1.55V and IBT is running at the moment. It seems that everything is ok as usual.

Any ideas what this could be? I am thinking about buying 2x8GB of 1600 RAM to replace my old set but the problem is that if it has nothing to do with the DIMMs, I wasted money because returning RAMs is not possible.

Click to expand...

If your going to purchase RAM, buy something a little faster than 1600 so you have head room to overclock with.

Look for DDR3-1866.

What effect does 1866 RAM have over 1600 regarding overclock?

cubusmedusa said:

What effect does 1866 RAM have over 1600 regarding overclock?

Click to expand...

A bit more stress on the IMC, otherwise, nothing.

If you have issues running 4 sticks then raise VCCIO/VTT and/or VCCSA voltage. Sometimes it's helping, sometimes not.
Command Rate 2 will be necessary while running 4 memory sticks on most boards.

Woomack said:

If you have issues running 4 sticks then raise VCCIO/VTT and/or VCCSA voltage. Sometimes it's helping, sometimes not.
Command Rate 2 will be necessary while running 4 memory sticks on most boards.

Click to expand...

I worked on Vccio and Vpll. I lowered Vpll to 1.6 and raised Vccio to 1.15. That allowed me to reduce Vcore to 1.38 which was impossible before. 5-6°C reduction in temperatures. The RAM seems to be fine sofar. I am still testing...