Having Serious Memory Issues

[Gorilla]

So....... to make a really long story short, I'm getting tons of blue screens. They usually give a memory related error (pagefault, memory error, etc).

Anyway, I've been getting these since I started using 8gb of memory instead of 2gb. Originally I thought I had a bad memory slot so I was only using 1 slot. Then newegg ran a sale.... so I bought another 2x2gb. Turns out that I just needed to up the memory voltage to 1.65 (though it's advertised as 1.5v) and the vcore.

Anyway, things ran well more or less for a week or so, but no longer.

I read somewhere that 1333 is technically the highest that is supported, so I thought maybe I could drop the ratio from 2:8 to 2:6 and take some strain of the controller. Well, that results in my system not POSTing. I just get an endless cycle of fans revving up and then stopping.

So, my next thought was to just drop my OC in favor of stability. Well, that failed. I dropped it down from 191 x 21 to 145 x 21 (3.05 or so) which dropped my memory at 2:8 to around 1165. My first startup attempt resulted in a blue screen before windows even finished loading.... so much for that idea.

Any suggestions?
My specs are in my sig.

 

[jubrany]

You need to make sure your RAM is stable before you overclock.

don't increase the memory voltage or core voltage.

Try running the system at stock BCLK, leave all that stuff on auto.

increase the QPI VTT. Try something like 1.25v.
manually key in 1.5v for the memory.

Manually key in your memory timings.... 9-9-9-24, 2N, and make sure tfrc is at least 88. Make tfaw at least 31. Let the rest be on Auto.
Set your Ram Frequency to 1600.

Should work. Throw 8 instances of HCI memtest at it, use 700-900 MB on each instance or whatever it takes to get most of the ram used up. If it's no good, it should fail fast, in under 2 hours likely.

If you can't get stable like that, set tRC to 40, tRRD to 5, tWR to 12, tWTR to 6 and tRTP to 6. If that still doesn't work, loosen trfc further to like 110 and tfaw to 39.

If that STILL doesn't work, start testing individual sticks with Memtest86+ v4.1.

I'm sure once you've tried all this you will have a stable system. Then you can start trying to OC

P.S. if after going through all this you still have problems, I would consider checking your processor mounting. Check the socket for foreign objects or bent pins. Check the underside of the CPU to see if all the pads have contact marks from the socket. Avoid overtightening the CPU cooler mounting.

 

[FunkZ]

Gskill rates the 2x2Gb kit at 9-9-9-24 but if you're using two kits you may have to loosen the timings some. It's always more taxing on the memory controller to address 4 sticks than 2.

 

[Gorilla]

Well, I tried relaxing the timing to 10-10-10-40 and completely eliminating my overclock. Didn't help. Also, it doesn't seem to be related to heat or anything. My cpu will be at 2% load and memory usage at 20% and I will still get blue screens when I'm not doing anything. Memory voltage is currently at 1.5v.

I haven't run memtest or anything recently, but I did when I first build this system and started having these problems and it ran fine. Maybe it's time to go back to 1 x2gb.

I kind of doubt that it had anything to do with the cpu seating since I only get problems when I go up past 2gb. If the processor wasn't making contact on some pins wouldn't I have issues all the time?

 

[Gorilla]

Updated my bios... that added some options that I haven't seen before, but windows blue screened every time I tried to start up until I went from the default memory settings to the relaxed 10-10-10-40 (they were at defaults because of the bios update).

Anyone want to tell me if CPU Skew Control is important? Because that is one of the new options.

 

[FunkZ]

I wouldn't think that should affect memory. After the BIOS flash you said you got BSOD every time until you tried the higher timings again. So is it booting to Windows now?

If still having problems might try running Memtest+ like jubrany suggested and see if errors show up, also at what block of memory it occurs, you might have 1 bad stick that only causes the error when Windows tries to use it.

 

[Gorilla]

Well I don't know. I completely eliminated my OC. I increased memory voltage to 1.65 (something G.Skill suggested even though it's rated at 1.5V). Lowered memory timings. But I still get blue screens. I'm sure that it isn't a bad stick because each one of them works just fine if I only use one stick at a time, but if I put more than 1 (2 or more) in I start to get blue screens. Maybe it's just the motherboard (this was my suspicion from the very start).

 

[jubrany]

^ Can you tell us what your QPI VTT and (if your board has it) QPI PLL is? these voltages are important to stability for your RAM.

 

[Gorilla]

^ Can you tell us what your QPI VTT and (if your board has it) QPI PLL is? these voltages are important to stability for your RAM.

I'm pretty sure all voltages except the memory voltage are set to auto. I used to have them set specifically, but I didn't see the point since I'm running everything at stock.

 

[jubrany]

I'm pretty sure all voltages except the memory voltage are set to auto. I used to have them set specifically, but I didn't see the point since I'm running everything at stock.

You are trying to run 4 sticks of memory, and running them at DDR3 1600 is still overclocking the memory.

This puts much much greater strain on the memory controller of your CPU.

I increased my QPI VTT even though I am only using 3 of my 6 memory slots. I needed to do this to stabilize my RAM. 1600MHz is hard to do otherwise.

Even 1333 is overclocking technically.

Have you tried the XMP profile??? Try that, and bump up your QPI voltage.

 

[Gorilla]

They're at 1333, not 1600. I'll trying bumping up the qpi. I don't think I have an xmp profile option.


EDIT: So by qpi voltage do you mean frequency? I used to have my qpi frequency at 4.270 GT/s (probably because that's where the EVGA overclocking guide for p55 motherboards set it). My only other option is 4.8 GT/s, so that's where I set it.

Any idea what MCH Strap is?

Edit 2: So I'm wondering about XMP. What is it? and..... when I bring up cpu-z, under the SPD tab I can select different memory slots. For slots 1 and 3, I see JEDEC #1,#2,#3,#4. But for slots 2 and 4 I see JEDEC #2, #3, #4, and XMP 1600.
And for slots 2 and 4 the SPD Ext. box has XMP listed, but for slots 1 and 3 the SPD Ext. box is blank.

 

[Gorilla]

Another update: I tried setting my memory timings to those shown under XMP-1600 in cpu-z. (9-9-9-25-40). That resulted in my system not POSTing until I pushed the reset cmos button. It will POST at 9-9-9-25-74 though.

 

[Gorilla]

Well, i tried running memtest again. Massive numbers of errors. So time to go through and figure out if it is a bad stick or a bad slot.

 

[jubrany]

try 9-9-9-24-88 and command rate to 2N. Make sure it's 2N!

QPI VTT is called CPU VTT on your board. That needs to be within 0.5v of your memory voltage for stability.

Your memory is rated for 1.5v. Don't bother jacking it up to 1.65. Go back to 1.5, and try 1.2v or 1.25v for your CPU VTT.

I bet you once you do that, things will be running a lot better.

 

[Gorilla]

try 9-9-9-24-88 and command rate to 2N. Make sure it's 2N!

QPI VTT is called CPU VTT on your board. That needs to be within 0.5v of your memory voltage for stability.

Your memory is rated for 1.5v. Don't bother jacking it up to 1.65. Go back to 1.5, and try 1.2v or 1.25v for your CPU VTT.

I bet you once you do that, things will be running a lot better.

Thanks for the advice. I'll check the cpu vtt later, but if iirc my cpu vtt used to be at 1.2V. Maybe 1.25v will make the difference.