Problem with F3-17000CL9D-8GBXLD and XMP

[avoyk1]

Hello!

I have been having some problems with my memory with respect to stability. After reading a everything I could find with similar issues on this and other forums, I started troubleshooting. here is what I found.

With XMP enabled the system BSODs and memtest86 throws up tonnes of errors. This still happened when I OC'd the CPU and even manually configured the memory settings in BIOS and upped voltages (VCCIO, DRAM, VCORE etc.).
Memtest86 passed both modules separately in both default settings and XMP enabled. At BIOS defaults (1600Mhz 1.5v) both modules either dual or single channel will work without an issue. But when XMP is enabled they and both modules are used, they are unstable in any config. Yesterday I updated BIOS to F10 and completely reconstructed my OC settings and turned on XMP, then ran Intel burn test on 'very high' setting for 10 runs and it passed oddly enough. Then I was able to run Prime95 for a couple hours without any issues of any sort. So I thought the BIOS update fixed it. But just to check I ran memtest86, and it threw up errors, though there were less than before.

So with XMP enabled the system currently appears to be stable, but memtest86 throws up errors. I'm not sure whether this is a memtest86 fault or if there is in fact a memory issue. Any and all insight would be greatly appreciated.

System specs:
AX1200 PSU
GA-Z86X-UD7 MOBO
G.SKILL F3-17000CL9D-8GBXLD RAM
INTEL I7 2600K CPU (OC 4.5GHZ)
H100 CPU COOLER
2X MSI NVIDIA GTX590 GFX CARDS
MISC. (SOUND CARD, HDDS, DVD BURNER)

[Tradesman]

Did you run both sticks together, if so that can create errors, should run each stick by itself. Might also check you have the latest version of Memtest, but from your test results with Prime, et al am guessing you did Memtest w/ both sticks at once

[avoyk1]

Thanks for the quick reply mate.

I did run both sticks at once, but memtest only had errors when XMP was enabled. Should this be disregarded?
Each stick individually with XMP doesn't give errors.
I think the BIOS update may have fixed the BSODs that used to occur with XMP.
So is this just a memtest issue, and there is no REAL problem with the RAM? I have the latest version of memtest86.

[Tradesman]

One of the advanced timings may be off a hair, could you post them and we'll take a look, I prob won't be back till later in PM but GSkill may be in before I'm back...from the sounds of it, it's prob Ok to run, may just be a BIOS glitch in the XMP implementation, have seen that lately

[avoyk1]

Timings XMP set are as follows (in order as they are in BIOS):

Channel A

Basic
9-11-9-28

Advanced
50-7-8-16-28-8-170-8-40-2

<Don't remember name this section>
2-39

Channel B (Pretty much the same as A)

Basic
9-11-9-28

Advanced
50-7-8-16-28-8-170-8-40-2

<Don't remember name this section>
1-39

I also did some more testing (Both sticks with XMP enabled).
I left the windows version of memtest running today while I was at school, came home and it had encountered errors that I'm guessing were about a couple hours or so into the test (test was at about 600% coverage) (screenshot below).
I then tried a custom run on Prime95 using about 6.5GB of ram. Windows BSOD's within seconds of starting the test. Error code 0x124, it mentioned something about kernel inpage error. So I tryed bumping QPI/VTT voltage from 1.050v to 1.080v and ran the same Prime95 test again, this time it ran smoothly for 3 hours until I stopped it (screenshot below). Now I'm currently running memtest for windows again to see what happens. 115% coverage and no errors yet.

Memtest when I got home (I had closed down the error messages that paused the test) (Ignore the voltages in HWmoniter, they are incorrect)


Prime 95 after about 3 hours (small FFT setting but with 6500MB memory)

[avoyk1]

*UPDATE*
Memtest is still giving me errors
...Continuing Prime testing on blend test but with 6500mb memory.

*EDIT*
Just after posting this, one of the Prime95 workers gave me the following error:

FATAL ERROR: Final result was 803F919D, expected: D1045A66.
Hardware failure detected, consult stress.txt file.

[avoyk1]

Also I just noticed how all these say different things... Whats the deal here... And shouldn't the 1069.8mhz in image 3 be something like 2133mhz??? I'm confused. BIOS says the frequency is about 2140mhz.

1

2

3

[Tradesman]

I'd try raising the DRAM voltage + .03 and up the QPI/VTT a bit more to stabilize, maybe + .02....Cpu-Z looks correct, showing 1069.8, the DRAM is DDR (Double Data Rate) so the 1069.8 (X 2) equates to 2139.6 (can read more about this in the thread I posted a while ago:

http://www.gskill.us/forum/showthread.php?t=10565

[avoyk1]

Ahh that makes sense now, thanks for that
I've tried raising the DRAM voltage to 1.7, and that caused it to fail to boot. Then I tried raising all manner of other things (within safe parameters) to try and stabilize, but the PC still wouldn't boot. Also tried raising/lowering Vcore and PLL to see if it had an effect (which it didn't).
Other ideas? Did the timings I posted seem to be correct (In post #5)?

[GSKILL TECH]

The timings are correct.

So the system boots up now but is unstable?

Thank you
GSKILL TECH

[avoyk1]

Yes it boots but is unstable.

[avoyk1]

Is this something to be concerned about? I don't know what 'hard faults' are but my guess is they are bad. This is with the memory running at default 1600mhz 1.5v (so I can play games without the PC going nuts).

[Tradesman]

It's basically like a page fault when a program looks for something in DRAM and it's not there so it has to read it from the disk, it generally happens most when multitasking (multiple programs/windows open of having a memory intensive program open, for most the more DRAM you have the less you'll see these as less will be written to disk.

[avoyk1]

Anything else we can do to try getting this working?

[Tradesman]

Hard faults, isn't really a good description, it's not really a fault so much as the system doesn't have what ir's looking for in DRAM, it's already moved it to disk to put other data in the DRAM, it isn't anything to worry about, there's more info here and a little more on monitoring:

http://technet.microsoft.com/en-us/l.../cc722173.aspx